Research ArticleAgent Based Modeling on OrganizationalDynamics of Terrorist Network
Bo Li Duoyong Sun Renqi Zhu and Ze Li
College of Information System and Management National University of Defense Technology Changsha 410072 China
Correspondence should be addressed to Bo Li libonudtgmailcom
Received 30 April 2015 Revised 31 August 2015 Accepted 11 October 2015
Academic Editor Aura Reggiani
Copyright copy 2015 Bo Li et al This is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Modeling organizational dynamics of terrorist network is a critical issue in computational analysis of terrorism research Thefirst step for effective counterterrorism and strategic intervention is to investigate how the terrorists operate with the relationalnetwork and what affects the performance In this paper we investigate the organizational dynamics by employing a computationalexperimentation methodology The hierarchical cellular network model and the organizational dynamics model are developedfor modeling the hybrid relational structure and complex operational processes respectively To intuitively elucidate this methodthe agent based modeling is used to simulate the terrorist network and test the performance in diverse scenarios Based on theexperimental results we show how the changes of operational environments affect the development of terrorist organization interms of its recovery and capacity to perform future tasks The potential strategies are also discussed which can be used to restrainthe activities of terrorists
1 Introduction
A key aspect of studying terrorist network is to investigatethe organizational dynamics and the mechanism by whichits structure grows and changes over time [1] Due to thenetworked connection between terrorists network analysismethods provide a powerful way to study the organizationalstructure The interplay between topology and dynamics incomplex networks is a fundamental problem which has notbeen widely studied [2] Individual behavior and networkties mutually influence each other [3] For studying terroristorganization it is important to find out how the terroristsbehave in performing tasks and how the activities affectthe organizational dynamics [4 5] On the other handterrorist network is usually not a complete picture due to theinherent difficulties in obtaining data It shows only thoselinks that have been publicly disclosed [6 7] Hence usingcomputational experimentation to obtain insight into thedilemma of terrorist network dynamics is valuable
Many reported empirical works have studied the struc-tural characteristics of terrorist network using Social NetworkAnalysis (SNA) [8ndash11] They provide both qualitative andquantitative results with empirical data and are helpful in
understanding the relational structure and finding key ele-ments (individual subgroup etc) However it is consensusthat terrorist organization is dynamic and far more compli-cated than structural network [12 13]Dynamic network anal-ysis (DNA) and metanetwork are used to model the complexorganizational structure and individual interaction [14] Inorder to analyze the operational and evolutional dynamics ofterrorist network agent based modelingsimulation (ABMS)is introduced to model the autonomous agents and theirinteractions [15ndash17] This method is effective in understand-ing the complicated evolution of system derived from agentinteraction with rich results of simulation experiments [1819] Lots of models are proposed to describe the terroristbehavior [20ndash22] Based on the ABMmethod changes detec-tion of organizational network characteristics [20] networkcoevolution of social and geospatial dimensions [22] anddynamics between networks [15] are investigated Thesereported works studied the terrorist network from differentperspectives and provided useful results to investigate howthe terrorists act and its influence on the organizationaldynamics
However there are two limitations in current reportedstudies Firstly the terrorist network should be analyzed
Hindawi Publishing CorporationDiscrete Dynamics in Nature and SocietyVolume 2015 Article ID 237809 17 pageshttpdxdoiorg1011552015237809
2 Discrete Dynamics in Nature and Society
Relation structure
The workflow
Organizationalrecovery
Organizational growth
Cellular network
Hierarchical command
The agentmodel
Interaction rules
Organizational dynamics driven
Organizationalmechanism
The organizationalprocess model
The organizationalstructural model
Agent based simulation framework Organizationaldynamics analysis
Environment(resource human)
Organizationalcost
Organizationalbenefit
Organizationalperformance
Number ofmembers
Number ofaction cells
Organizationalgrowth
Statisticalresult
Figure 1 The systematic framework of the research on organizational dynamics of terrorist network
with the dynamic organizational process Terrorists takeactions based on not only the relational network but alsothe established operational process The relational networkevolves under different conditions and makes the analysis ofSNA ineffective [23] The other limitation is that the mecha-nism of recovery in studying the dynamics or interventionstrategies has not been fully considered The resilience ofterrorist network provides the organization with the ability ofsurvival and growth In fact understanding how to suppressthe recovery of terrorist network is more important thanthe identification of key nodes for developing strategies ofcounterterrorism [24]
In this paper we study the organizational dynamics ofterrorist network by employing the computational experi-mentation method The rest of the paper is organized asfollows Section 2 describes the overall research frameworkof this paper Section 3 gives the model of hierarchicalcellular network to describe the functional terrorist networkThe actions based on this organizational structure are alsodefined Section 4 proposed the organizational processmodelto describe the organizational dynamics The mechanisms oftask workflows organizational recovery and organizationalgrowth are proposed to model the operational processes Inaddition the resilience of terrorist network is incorporatedinto the model by the recovery of organizational elementsSection 5 gives the evaluations of organizational perfor-mance Both costs and benefits are included to analyze theperformance In Section 6 by employing the ABM methodthe terrorist network is simulated in both open and limitedenvironments and the impacts of different factors are testedunder various dynamic scenarios With the results we showhow the terrorist network works as an organization and whatelements affect the performance The potential strategies are
also discussed which exhibits the potential applications forpublic security department Section 7 concludes the paper
2 Research Framework
In order to study the organizational dynamics of terroristnetwork the agent based modeling is employed to constructthe computational framework of the terrorist organizationAs shown in Figure 1 the framework consists of four compo-nents (1)The agent based simulation framework is used tomodel the complex interaction between the terrorists withinthe organization (2)The organizational structural model isused to describe the relation between the terrorists and thecorresponding command structure (3) The organizationalprocess model includes the workflow the organizationalmechanisms of recovery and growth This model providesthe organizational driven and behavioral mechanism (4)Theorganizational dynamics analysis includes the organizationalperformance and growthThe performance reflects the func-tional dynamics during the organizational activities and thegrowth represents the structural dynamics
The ABS framework provides the tool to model the com-plex interaction behavior between terrorists and to observethe organizational dynamics The organizational models areadded to the framework as the elements of agent basedmodeling on terrorist organization Cellular network andhierarchical command relation are used to describe the spe-cific structure and relation in terrorist network The processmodel consists of threemechanisms the workflow the recov-ery and the growth The workflow generates organizationaldriven and individual behaviors and it gives the generalrules of organizational activities The recovery mechanismand the growth mechanism give two rules of organizational
Discrete Dynamics in Nature and Society 3
Conductcell
Action cell
Recruit cellResource cell
Training cellHumansupplement
Humansupplement
Human supplement
Resourcesupplement
Action cellAction
cell
x1
x2
x3
x4
x5
x6
x7
x8
x9
x10
x11
x12
x13
x14
x14
x15
x16
Figure 2 Organizational structural model of terrorist network
activities in specific situations As the simulation time goeson the statistical results are changing and the organizationaldynamics analysis is carried out based on the results capturedin the system
3 The Organizational Structural Model ofTerrorist Network
Terrorist network is different from social network in twoaspects (a) cellular groups as basic activity units and (b)hierarchical command structure Terrorists operate accord-ing to this hybrid structure which makes the network do notcorrespond to a particular kind of networkmodel [25] Basedon the description of terrorist network in [26] a theoreticalterrorist network is shown in Figure 2The characteristics canbe understood from three aspects the cellular network struc-ture the hierarchical command structure and the individualactions based on the network
31 Cellular Network Structure It has been reported thatcellular network is the most suitable structure to describeterrorist organization [27 28] It usually consists of manycell groups (or subgroups) which are the basic active unitsto perform tasks [29 30] As shown in Figure 2 the cellsare cohesive structure of a set of connected individuals whocooperate to perform the task The reason for terrorists toadopt this network topology is structural compartmentaliza-tion [26] which is a trade-off between efficiency and security
Different network structures are adopted by the cellsto fulfil the diverse functional requirements [31] In thismodel we hypothesize that all the cellsrsquo structures conform tocorresponding network models Action cells have the struc-ture of fully connected network for frequent communicationbetween the members in it and so is the conduct cell Thestructures of recruit cells and resource cells conform to starnetwork which are helpful for the cells to perform task
(resource acquirement and human recruitment) widely andcovertlyWe also use star network here to depict the structureof training cell because of the function of the cell trainingthe terrorists singly as a ldquofactoryrdquo where the terrorists arethe products and are transferred to the other cells for humansupplement
The terrorist network can be regarded as a graph 119866 =(119873 119864) where 119873 represents the set of agents (terrorists)and 119864 is the set of all the relations in the network Wehypothesize that the network only includes the member ofthe organization whichmeans the agents in the environmentare not considered Only the formal relations are included inthis model and the informal relations such as social relationare not considered in this paper The cells can be regarded assubgroups of119866 In thismodel five types of cell are consideredaccording to their function in the organization and thenotations are listed inTable 1 aswell as the explanations Eachcell contains a node set (as 119873AC in 119866AC) and an edge set (as119864AC in 119866AC) which is the set of all the relations between thenodes in the cell As there is more than one action cell we use119899ac to represent the number of action cells
32 Hierarchical Command Structure
321 Role of Terrorist and Type of Relation Recall that terror-ist network is different from social network in organizationalactivities As an organization the roles of terrorists are diverseand the relational type depends on their roles Based on thecellular network structure we define two types of roles in themodel the cell leader and the cell memberThe role is definedas a property of agent and the definitions are as follows
Cell leader (denoted as Role119897) is the leader of a cell Itorganizes the activities of the cell members and connects tothe superior cell for organizational command
Cell member (denoted as Role119898) is the member of a cellIt performs the specific activities in task process
We also define two types of relations in this model thepeer-to-peer relation and the superior-subordinate relation
Peer-to-peer relation (denoted as Relpp) is the type of linkbetween the agents with the same role
Superior-subordinate relation (denoted as Relss) is thetype of link between the agents with different roles
As the terrorist organization has hierarchical structurewe give the assumptions used in this model
(1) The type of the relation between the members in thesame cell is Relpp and the type of the relation betweenthem and the cell leader is Relss
(2) The member of the conduct cell is the superior of theother cells which means the relations between themand the leaders of the other cells are Relss
(3) There is no relation between the cells leaders that isa cell leader cannot interact directly with another cellleader
These assumptions are given based on the characteristicsof terrorist organization and empirical research According tothe definition and described cellular network in Section 31the hierarchical structure of terrorist network is fixed to some
4 Discrete Dynamics in Nature and Society
Table 1 The notations of the cells in the cellular network
Notation Explanations119866AC Action cell and 119866AC = (119873AC 119864AC) sub 119866119866RC Resource cell and 119866RC = (119873RC 119864RC) sub 119866119866Co Conduct cell and 119866Co = (119873Co 119864Co) sub 119866119866TC Training cell and 119866TC = (119873TC 119864TC) sub 119866119866Re Recruit cell and 119866Re = (119873Re 119864Re) sub 119866119866119894
AC The 119894th action cell and 119866119894AC = (119873119894
AC 119864119894
AC) sub 119866
extent The conduct cell is barely changed and the roles ofthe members are specified The leaders in the other cells aredesignated by the superior in the conduct cell which is tosome extent fixed as well
322 Intercell Hierarchical Structure The command relationbetween cells represents inherently hierarchies between func-tional compartments As illustrated in Figure 2 the tasks areoperated by the conduct cell with the instructions sent tothe other cells Firstly the conduct cell sends missions tothe action cells that then take actions required to performthe tasks When the conduct cell receives the feedback ofhuman and resource requirements it passes human andresource supplements to the training cell and resource cellseparately This hierarchical command structure makes thecore of organization able to coordinate the members for anoverall goal and maintain their own agendas Note that theorders are transmitted by the individuals in cells because thecell is not a real entity but an organizational form of groupmembers
323 Intracell Hierarchical Structure Decentralization attactical level is due to the difficulty of real-time commandand control within a large clandestine cellular network [26]Cell leaders usually have freedom of tactical decision-makingand action based on local condition which produces intracellhierarchies and self-organization As shown in Figure 2leaders of functional cells (1199095 1199096 11990910) are receivers ofthe assigned missions They organize the actions of othermembers in cells to complete tasks Take action cell (where 1199095stays) as an example 1199095 receives a mission from the superiorcell (conduct cell) then it evaluates the requirements basedon the condition of the action cell and sends the feedback tothe superior cell During the action period 11990911 is in charge ofreceiving resource from the resource cell and acts under thecommand of 1199095 because it is the only one that communicateswith the superior cell Even in the core cell the members(1199092 1199093 1199094) are under the command of the cell leader (1199091) whois either the leader of the organization or the only one thatconnects to the superior leader
4 The Organizational Process Model ofTerrorist Network
The organizational process model is proposed to describe theoperational mechanism of terrorist network First the agentmodel is given and the actions are defined as a set of available
actions Then the response of agent in different mechanismis givenThe organizational processmodel can be understoodas the interactional rule of agent [17]
41 The Agent Model The system considers a populationof agents (terrorists) that continually act as the model pro-gresses As a terrorist organization the individual behaviorsand attributes are complicated For clarity we only considerthe following attributes which are related to the organiza-tional dynamics in this model
Role This is the role of agent in the cell At the beginningthe leader of a cell is assigned and can be identified by theindividual structure Once the leader of a cell is lost a newleader will be generated in the cell and this will be discussedin the recovery mechanism
Age This is the length of time that an agent stays in thenetwork
Resource This is the attribute which re-presents if an agentholds resource
Cell Type This is the type of the cell which the agent belongsto
The actions are limited by the role and the cell typeof agent which means that agents have different candidateaction set according to their position in the organizationTheactions used in the model are defined as follows
Requirement PassingThe requirement for resource or humanis passed from the source agent to the target agent Thisbehavior is defined as ReP119894119895 which represents that therequirement is passed from agent 119894 to agent 119895 The cost ofpassing a requirement is CostReP
Resource Transferring The required resource is transferredby terrorists through established channels (solid arrows inFigure 2) The element transferring behavior is defined asTra119894119895 where 119894 and 119895 are the terrorists transferring the element
Moving The agent moves from one cell to another cellfor human supplement This behavior is denoted as Mo119894and 119894 is the agent There are some differences between theresource transferring and the moving Agent remains in theresource cell and it links to the required cell after resource istransferred Once an agent moves to a new cell it would notconnect to the original cell whichmeans the other cell cannotdirectly send the human requirement to the training cell butit can send the resource requirement to the resource cell aslong as the link exists
Building a Link This action is taken when a link is neededbetween two agents and it is denoted as BuL119894119895 where 119894 and119895 are the agents who built the relation As there are two typesof links the newly built link depends on the situation andthe roles of the two agents which will be discussed in thefollowing organizational mechanism
Discrete Dynamics in Nature and Society 5
Table 2 The candidate action set of an agent in terrorist network
Role119897
Role119898
Action cell ReP BuL ReP Tra BuLResource cell ReP BuL Res ReP Tra BuL ResConduct cell ReP ReP BuLTraining cell ReP BuL Mo BuL Train RePRecruit cell ReP BuL Rec Mo BuL Rec ReP
Recruitment Recruitment (denoted as Rec) is performed bythe members of recruit cell and it is the only way to generatenew member of terrorist network We use 119875Recruit to denotethe probability of successful recruitment and Costrecruit torepresent the cost of one recruit
Training The new members will take certain time in trainingbefore they are assigned to a cell and this action is denotedas Train119894 The time of training is denoted as 120590Training and thecost is Costtraining
Resource Acquirement The resources are acquired by themembers of resource cell We assume that the resourcein this model contains kinds of needs for performing atask The probability of resource acquirement is denoted as119875Resource and the cost is denoted as Costresource This action isrepresented as Res
The candidate action set of an agent is shown in Table 2
42 Task Workflow The purpose of terrorist organization isto produce terror events as much as possible and we simplifythe event to a task in this model The requirement of a taskis human and resource and they are simplified as agentsand resource The general workflow of terror organization ismodeled as the following four steps
Step 1 (generating task) The task is generated by the cellleader of action cell If an action cell is not operating anytask a new task will be created by the cell leader Eachtask 120593 requires amount of resource and agents whichcan be denoted as 120593(resource) and 120593(agent) respectivelyrepresenting the number of resources and agents required tofulfill a task
Step 2 (generating requirement) The requirement is evalu-ated by the cell leader and the element requirements will besent by the agents in the cell Assuming the size of action cell119866119894
AC is |119873119894AC| then the required number of agents is
119873Requirement = 120593 (agent) minus10038161003816100381610038161003816119873119894
AC10038161003816100381610038161003816 (1)
where119873Requirement is the number of generated human require-ments The number of required resources is 120593(resource) Theprocesses of human supplement and resource supplement aredifferent The human requirement will be directly sent to thesuperior and for the resource the leader will interact with allthemembers If there is a member connecting to the resource
cell the requirement will be sent to the member otherwiseit will be sent to the superior
Step 3 (agent response) After the requirements are sent theagent who receives the message will process the requirementbased on its role and relation Figure 3 shows the responsemechanism of an agent to a requirement The agent interactswith the other agent and it can get the information about theother agent In this response process the type of the link builtbetween two members in different cells is Relpp
Step 4 (task execution) Once the requirements are fulfilledthe action cell will launch an event and this is done by thecell leader of action cell The activity circle of action cell canbe illustrated as Figure 4 We hypothesize that the memberswill be lost in performing the task The probability for eachmember to be arrested is denoted as 119875arrest which dependson the efficiency of counterterror department
43 Organizational Recovery Mechanism The recovery pro-cess of terrorist network is the key mechanism for organiza-tion to survive Resilience is a dynamic process associatedwith systems that persist and perform their primary tasksunder pressure from exogenous shocks [32] We do notconsider the case of large-scale attack to the terrorist networkand focus on the normal consumption during the operationalprocess (a member can also be lost if he changes his mindwhich may happen because of the decay of his faith inthe organization In order to focus on the organizationaldynamics we hypothesize that the members will not be lostbecause of the belief issue) The loss in the task concernstwo types of cells the action cell and the resource cell Therecovery process is triggered after a task is performed Ourhypothesis is that the recovery mechanism comprises twobehaviors
(1) Check the leader of the action cell If the leader ofthe action cell is lost a new leader will be generatedfrom the remaining members The selection rule isbased on the attribute Age that is the agent whostays the longest in the cell will be designated as thenew leader This is because when an individual stayslonger in the organization it usually has the richestexperience (expertise knowledge etc) which is themost important characteristic of an action leader
(2) After that the new leader tries to build a new linkto the superior cell with the probability 119875reconnect Thetype of the newly built link is RelssThen the recoveryaction is finished
Figure 5 displays the overall recovery process of anaction cell It can be found that the recovery mechanismof the organization is partly finished by the organizationaloperational process
Resource cell also has recovery mechanism as the mem-ber may be lost in the process of performing a task In themodel we only consider the operational consumption andhypothesize that the member is only lost when the agent in
6 Discrete Dynamics in Nature and Society
If it is a member ofrequired cell
Find a neighborthat connects to the
required cell
Pass the requirement tothe agent
No
Yes
Pass the requirement tothe superior
No
If it can fulfil therequirement
Yes
Resource transferring
If it connects to therequiring cell
Yes
Yes
Find a member in thecell that can fulfilthe requirement
No
No
YesBuild a link to the select
agent in the requiring cell
No
Type of requirement
Resource
Human moving
Human
Receive a requirement
Finish a requirement
ActionChoice
Figure 3 The response mechanism of agent to a requirement
Generaterequirements
Fulfil therequirements Perform a task
Recovery Human loss
No loss
Figure 4 The activity circle of action cell
action cell who links to it is arrestedThe recoverymechanismof resource cell is triggered by the event that a member islost When a member is lost then a human requirement isgenerated by the cell leader
As illustrated in Figure 6 if member 1199096 who is in theaction cell and communicates with the resource sender 1199095 isarrested then 1199095 is lost tooThe leader of resource cell 1199091 then
sends human requirement to the conduct cell for supplementso as tomaintain functional completeness Even though somemembers are lost the resource cell can also acquire resourcebut the efficiency is lower than the complete cell
44 Organizational GrowthMechanism The launched eventslead to organizational growth and the correspondingincrease in size leads to faster production of new events[33] The more frequent the terrorist acts the faster thenetwork grows Events and social influence promote theappeal of organization and recruitment of new militants [12]For this consideration we assume that new action cell willbe generated after certain tasks are finished and a threshold120590Task is used as the trigger of the organizational growth Thesteps of growth can be modeled as follows
Step 1 (leader requirement) A human requirement is gener-ated and assigned to the member in the conduct cell who isin charge of action cell
Discrete Dynamics in Nature and Society 7
Lose members
Lose leader Generate new leader Human supplement
Human supplement New action cell
New action cell
Reconnection
Action cellPerform task
Perform task
Cell leaderx1
x1
x1
x1
x2
x2
x2
x2
x2
x3
x3x
3x3
x3
x3
x3
x3
x4
x5
x5
x5
x5
x5
x5 x
5
x5
x6
x6
x7
x7
x8
x8
x9
x9
Figure 5 The recovery process of action cell
Resourcesupplement
Action cell
Memberlost
Action cell
Resource cell Lose member Human supplement New resource cell
x1
x1x
1
x2
x2x
2
x3
x3x
3
x4
x4
x4
x5
x6
x1
x2
x3
x4
x5
x6
x7
x7
Figure 6 The recovery mechanism of resource cell
Step 2 (leader designation) After the target agent moving tothe new cell it is assigned as the leader of the new actioncell In this process the role of the agent is changed from amember of the training cell to the leader of the new cell
Step 3 (cell formation) Although the newly built action cellhas only one agent it can be performed as a cell whichmeansit can be assigned a task and generate the human and resourcerequirements and send it to the other agents The completeaction cell would form across the organizational task flow
The mechanism of organizational growth is shown inFigure 7 When the tasks performed by action cells reacha threshold 120590Task the network grows and establishes a newaction cell First the leader of the new cell (1199092) is assignedby the operational leader (1199091) in conduct cell Secondlyrequirement of human supplement is sent by the new leaderThirdly the members are transferred through the workflowof human supplement until the new action cell is formedSimultaneously the sizes of resource cell recruitment celland training cell all grow in order to satisfy the growingdemand of actions
5 Organizational Performance Evaluations
In order to analyze the organizational dynamics the indica-tors for evaluating performance are constructed based on thestructural model described in Section 3
As mentioned in [34] we evaluate the performance usingcost and benefit The cost consists of CostReP CostTrainingCostRecruit and CostResource These costs can be recordedduring the simulation and the total cost Costtotal can becalculated as
Costtotal = CostReP + CostTraining + CostRecruit
+ CostResource(2)
Organizational Size The most intuitional evaluation of thedevelopment of terrorist network is the organizational sizethat is the total number of agents in the terrorist organiza-tion and it can be calculated as follows
Size119866 = |119866|
=1003816100381610038161003816119866RC1003816100381610038161003816 +1003816100381610038161003816119866Co1003816100381610038161003816 +1003816100381610038161003816119866TC1003816100381610038161003816 +1003816100381610038161003816119866Re1003816100381610038161003816 + sum119894
10038161003816100381610038161003816119866119894
AC10038161003816100381610038161003816
(3)
Action CellsThe number of action cells measures the growthof terrorist organization from the operational aspectThis canbe measured through the value of 119899ac
Organizational Benefits The goal of terrorist organization isto perform attacks or take similar activities to reach politicalpurpose It is intuitional to use the number of completed tasksas the organizational benefits The task set Φ contains all the
8 Discrete Dynamics in Nature and Society
Conduct cell
Action leader
Conduct cell
Leader of newaction cell
Conduct cell Conduct cell
Conduct cell Assign leader Human supplement Generate new action cell
x1
x1x
1
x1
x2
x2
x2
x3
x3
x4
x4
x5
x6
Figure 7 The growth mechanism of terrorist network
tasks that have been performed The benefit can be obtainedusing
Benefit = |Φ| = 119896 (4)
Cost-Benefits Ratio This is to evaluate the cost-benefit of theactivity of performing tasks As the terrorist organizationalactivity is task driven we here only consider the requirementpassing behavior which measures the efficiency of how theterrorist is organizedThe cost-benefit is measured by the rateof cost and benefit which is denoted as CB so the cost-benefitof requirement passing CBReP is calculated as
CBReP =CostRePBenefit (5)
The other cost-benefits CBTraining CBRecruit andCBResource can also be calculated as (5)
6 Experimental Results and Discussions
In order to quantify the organizational dynamics we per-formed the experiments in a simulation environment anddiscussed the performance and the influence factors basedon the results Potential strategies for counterterrorism arediscussed with the results of sensitivity analysis
61 Experimental Setup The initial organizational structureof terrorist network was generated as shown in Figure 2 aswell as the roles of the individuals We tested the organiza-tional dynamics model in an open environment where thenumber of actions at each simulation step was not limitedIn order to carry out a comparative analysis we also useda limited environment with a parameter to represent thelimited cost 120590Cost which was the number of actions that canbe taken at each step in the network
(1) The initial terrorist network included a conduct cellthree action cells a resource cell a recruit cell and atraining cell The simulation step was 1000
(2) At each step cell leaders checked the situation of thecells and took actions based on the action set andworkflows (eg action cell performs the task if allthe requirements are finished) After that themessagelists of the agents who related to the actions wereupdated
(3) The structure of terrorist network was changed basedon the results of the cell actions (lose members linksetc)
(4) At each step the agents updated their message listsaccording to the interactive mechanism describedabove including message lists of themselves and themessage lists of the interactive objectsThe number ofactions which can be taken by an agent at one step isdenoted as 120590Behavior
(5) At each step the organization checks the number ofcompleted tasks If it reached a threshold 120590Task a newaction cell would be generated by the member in theconduct cell This means that a new action cell isgenerated at the time every 120590Task tasks are performed
(6) In the limited environments if the amount of actionsat each step reached a threshold 120590Cost the actions ofthe rest of agents at current step were stopped untilthe next step
For clarity the values of parameters that are used in thisexperiment and their explanations are listed in Table 3
62 Organizational Dynamics
621 Organizational Performance Figure 8 shows the sim-ulation results of organizational performance First theincrease of costs (Figure 8(a)) and tasks (Figure 8(b)) tendsto accelerate over time and this is because of the growth ofterrorist organization This is consistent with the results ofempirical studies that the violent events tend to acceleratewith increasing size and experience [33] There is a similarlinear relationship between the cost and completed tasksAlthough the cost during the processes of recruiting andresource acquirement depends on the probabilities whichare influenced by randomness the results show statisticallinear relationship It is easy to understand because theorganizational behaviors are driven by task Second thevalues of CostRecruit and CostResource are much higher thanCostTraining and this is caused by the probabilistic behav-iors of recruitment and resource acquirement The sum ofCostTraining CostRecruit and CostResource is much higher thanthat of CostReP and this is consistent with the results ofempirical studies that the preparation of a task is much moreexpensive than performing a task because multiple actionsare required to launch an attack Third Figure 8(a) showsthe differences caused by the task workflows CostRecruit andCostResource are influenced by the probabilities The action
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Discrete Dynamics in Nature and Society
Relation structure
The workflow
Organizationalrecovery
Organizational growth
Cellular network
Hierarchical command
The agentmodel
Interaction rules
Organizational dynamics driven
Organizationalmechanism
The organizationalprocess model
The organizationalstructural model
Agent based simulation framework Organizationaldynamics analysis
Environment(resource human)
Organizationalcost
Organizationalbenefit
Organizationalperformance
Number ofmembers
Number ofaction cells
Organizationalgrowth
Statisticalresult
Figure 1 The systematic framework of the research on organizational dynamics of terrorist network
with the dynamic organizational process Terrorists takeactions based on not only the relational network but alsothe established operational process The relational networkevolves under different conditions and makes the analysis ofSNA ineffective [23] The other limitation is that the mecha-nism of recovery in studying the dynamics or interventionstrategies has not been fully considered The resilience ofterrorist network provides the organization with the ability ofsurvival and growth In fact understanding how to suppressthe recovery of terrorist network is more important thanthe identification of key nodes for developing strategies ofcounterterrorism [24]
In this paper we study the organizational dynamics ofterrorist network by employing the computational experi-mentation method The rest of the paper is organized asfollows Section 2 describes the overall research frameworkof this paper Section 3 gives the model of hierarchicalcellular network to describe the functional terrorist networkThe actions based on this organizational structure are alsodefined Section 4 proposed the organizational processmodelto describe the organizational dynamics The mechanisms oftask workflows organizational recovery and organizationalgrowth are proposed to model the operational processes Inaddition the resilience of terrorist network is incorporatedinto the model by the recovery of organizational elementsSection 5 gives the evaluations of organizational perfor-mance Both costs and benefits are included to analyze theperformance In Section 6 by employing the ABM methodthe terrorist network is simulated in both open and limitedenvironments and the impacts of different factors are testedunder various dynamic scenarios With the results we showhow the terrorist network works as an organization and whatelements affect the performance The potential strategies are
also discussed which exhibits the potential applications forpublic security department Section 7 concludes the paper
2 Research Framework
In order to study the organizational dynamics of terroristnetwork the agent based modeling is employed to constructthe computational framework of the terrorist organizationAs shown in Figure 1 the framework consists of four compo-nents (1)The agent based simulation framework is used tomodel the complex interaction between the terrorists withinthe organization (2)The organizational structural model isused to describe the relation between the terrorists and thecorresponding command structure (3) The organizationalprocess model includes the workflow the organizationalmechanisms of recovery and growth This model providesthe organizational driven and behavioral mechanism (4)Theorganizational dynamics analysis includes the organizationalperformance and growthThe performance reflects the func-tional dynamics during the organizational activities and thegrowth represents the structural dynamics
The ABS framework provides the tool to model the com-plex interaction behavior between terrorists and to observethe organizational dynamics The organizational models areadded to the framework as the elements of agent basedmodeling on terrorist organization Cellular network andhierarchical command relation are used to describe the spe-cific structure and relation in terrorist network The processmodel consists of threemechanisms the workflow the recov-ery and the growth The workflow generates organizationaldriven and individual behaviors and it gives the generalrules of organizational activities The recovery mechanismand the growth mechanism give two rules of organizational
Discrete Dynamics in Nature and Society 3
Conductcell
Action cell
Recruit cellResource cell
Training cellHumansupplement
Humansupplement
Human supplement
Resourcesupplement
Action cellAction
cell
x1
x2
x3
x4
x5
x6
x7
x8
x9
x10
x11
x12
x13
x14
x14
x15
x16
Figure 2 Organizational structural model of terrorist network
activities in specific situations As the simulation time goeson the statistical results are changing and the organizationaldynamics analysis is carried out based on the results capturedin the system
3 The Organizational Structural Model ofTerrorist Network
Terrorist network is different from social network in twoaspects (a) cellular groups as basic activity units and (b)hierarchical command structure Terrorists operate accord-ing to this hybrid structure which makes the network do notcorrespond to a particular kind of networkmodel [25] Basedon the description of terrorist network in [26] a theoreticalterrorist network is shown in Figure 2The characteristics canbe understood from three aspects the cellular network struc-ture the hierarchical command structure and the individualactions based on the network
31 Cellular Network Structure It has been reported thatcellular network is the most suitable structure to describeterrorist organization [27 28] It usually consists of manycell groups (or subgroups) which are the basic active unitsto perform tasks [29 30] As shown in Figure 2 the cellsare cohesive structure of a set of connected individuals whocooperate to perform the task The reason for terrorists toadopt this network topology is structural compartmentaliza-tion [26] which is a trade-off between efficiency and security
Different network structures are adopted by the cellsto fulfil the diverse functional requirements [31] In thismodel we hypothesize that all the cellsrsquo structures conform tocorresponding network models Action cells have the struc-ture of fully connected network for frequent communicationbetween the members in it and so is the conduct cell Thestructures of recruit cells and resource cells conform to starnetwork which are helpful for the cells to perform task
(resource acquirement and human recruitment) widely andcovertlyWe also use star network here to depict the structureof training cell because of the function of the cell trainingthe terrorists singly as a ldquofactoryrdquo where the terrorists arethe products and are transferred to the other cells for humansupplement
The terrorist network can be regarded as a graph 119866 =(119873 119864) where 119873 represents the set of agents (terrorists)and 119864 is the set of all the relations in the network Wehypothesize that the network only includes the member ofthe organization whichmeans the agents in the environmentare not considered Only the formal relations are included inthis model and the informal relations such as social relationare not considered in this paper The cells can be regarded assubgroups of119866 In thismodel five types of cell are consideredaccording to their function in the organization and thenotations are listed inTable 1 aswell as the explanations Eachcell contains a node set (as 119873AC in 119866AC) and an edge set (as119864AC in 119866AC) which is the set of all the relations between thenodes in the cell As there is more than one action cell we use119899ac to represent the number of action cells
32 Hierarchical Command Structure
321 Role of Terrorist and Type of Relation Recall that terror-ist network is different from social network in organizationalactivities As an organization the roles of terrorists are diverseand the relational type depends on their roles Based on thecellular network structure we define two types of roles in themodel the cell leader and the cell memberThe role is definedas a property of agent and the definitions are as follows
Cell leader (denoted as Role119897) is the leader of a cell Itorganizes the activities of the cell members and connects tothe superior cell for organizational command
Cell member (denoted as Role119898) is the member of a cellIt performs the specific activities in task process
We also define two types of relations in this model thepeer-to-peer relation and the superior-subordinate relation
Peer-to-peer relation (denoted as Relpp) is the type of linkbetween the agents with the same role
Superior-subordinate relation (denoted as Relss) is thetype of link between the agents with different roles
As the terrorist organization has hierarchical structurewe give the assumptions used in this model
(1) The type of the relation between the members in thesame cell is Relpp and the type of the relation betweenthem and the cell leader is Relss
(2) The member of the conduct cell is the superior of theother cells which means the relations between themand the leaders of the other cells are Relss
(3) There is no relation between the cells leaders that isa cell leader cannot interact directly with another cellleader
These assumptions are given based on the characteristicsof terrorist organization and empirical research According tothe definition and described cellular network in Section 31the hierarchical structure of terrorist network is fixed to some
4 Discrete Dynamics in Nature and Society
Table 1 The notations of the cells in the cellular network
Notation Explanations119866AC Action cell and 119866AC = (119873AC 119864AC) sub 119866119866RC Resource cell and 119866RC = (119873RC 119864RC) sub 119866119866Co Conduct cell and 119866Co = (119873Co 119864Co) sub 119866119866TC Training cell and 119866TC = (119873TC 119864TC) sub 119866119866Re Recruit cell and 119866Re = (119873Re 119864Re) sub 119866119866119894
AC The 119894th action cell and 119866119894AC = (119873119894
AC 119864119894
AC) sub 119866
extent The conduct cell is barely changed and the roles ofthe members are specified The leaders in the other cells aredesignated by the superior in the conduct cell which is tosome extent fixed as well
322 Intercell Hierarchical Structure The command relationbetween cells represents inherently hierarchies between func-tional compartments As illustrated in Figure 2 the tasks areoperated by the conduct cell with the instructions sent tothe other cells Firstly the conduct cell sends missions tothe action cells that then take actions required to performthe tasks When the conduct cell receives the feedback ofhuman and resource requirements it passes human andresource supplements to the training cell and resource cellseparately This hierarchical command structure makes thecore of organization able to coordinate the members for anoverall goal and maintain their own agendas Note that theorders are transmitted by the individuals in cells because thecell is not a real entity but an organizational form of groupmembers
323 Intracell Hierarchical Structure Decentralization attactical level is due to the difficulty of real-time commandand control within a large clandestine cellular network [26]Cell leaders usually have freedom of tactical decision-makingand action based on local condition which produces intracellhierarchies and self-organization As shown in Figure 2leaders of functional cells (1199095 1199096 11990910) are receivers ofthe assigned missions They organize the actions of othermembers in cells to complete tasks Take action cell (where 1199095stays) as an example 1199095 receives a mission from the superiorcell (conduct cell) then it evaluates the requirements basedon the condition of the action cell and sends the feedback tothe superior cell During the action period 11990911 is in charge ofreceiving resource from the resource cell and acts under thecommand of 1199095 because it is the only one that communicateswith the superior cell Even in the core cell the members(1199092 1199093 1199094) are under the command of the cell leader (1199091) whois either the leader of the organization or the only one thatconnects to the superior leader
4 The Organizational Process Model ofTerrorist Network
The organizational process model is proposed to describe theoperational mechanism of terrorist network First the agentmodel is given and the actions are defined as a set of available
actions Then the response of agent in different mechanismis givenThe organizational processmodel can be understoodas the interactional rule of agent [17]
41 The Agent Model The system considers a populationof agents (terrorists) that continually act as the model pro-gresses As a terrorist organization the individual behaviorsand attributes are complicated For clarity we only considerthe following attributes which are related to the organiza-tional dynamics in this model
Role This is the role of agent in the cell At the beginningthe leader of a cell is assigned and can be identified by theindividual structure Once the leader of a cell is lost a newleader will be generated in the cell and this will be discussedin the recovery mechanism
Age This is the length of time that an agent stays in thenetwork
Resource This is the attribute which re-presents if an agentholds resource
Cell Type This is the type of the cell which the agent belongsto
The actions are limited by the role and the cell typeof agent which means that agents have different candidateaction set according to their position in the organizationTheactions used in the model are defined as follows
Requirement PassingThe requirement for resource or humanis passed from the source agent to the target agent Thisbehavior is defined as ReP119894119895 which represents that therequirement is passed from agent 119894 to agent 119895 The cost ofpassing a requirement is CostReP
Resource Transferring The required resource is transferredby terrorists through established channels (solid arrows inFigure 2) The element transferring behavior is defined asTra119894119895 where 119894 and 119895 are the terrorists transferring the element
Moving The agent moves from one cell to another cellfor human supplement This behavior is denoted as Mo119894and 119894 is the agent There are some differences between theresource transferring and the moving Agent remains in theresource cell and it links to the required cell after resource istransferred Once an agent moves to a new cell it would notconnect to the original cell whichmeans the other cell cannotdirectly send the human requirement to the training cell butit can send the resource requirement to the resource cell aslong as the link exists
Building a Link This action is taken when a link is neededbetween two agents and it is denoted as BuL119894119895 where 119894 and119895 are the agents who built the relation As there are two typesof links the newly built link depends on the situation andthe roles of the two agents which will be discussed in thefollowing organizational mechanism
Discrete Dynamics in Nature and Society 5
Table 2 The candidate action set of an agent in terrorist network
Role119897
Role119898
Action cell ReP BuL ReP Tra BuLResource cell ReP BuL Res ReP Tra BuL ResConduct cell ReP ReP BuLTraining cell ReP BuL Mo BuL Train RePRecruit cell ReP BuL Rec Mo BuL Rec ReP
Recruitment Recruitment (denoted as Rec) is performed bythe members of recruit cell and it is the only way to generatenew member of terrorist network We use 119875Recruit to denotethe probability of successful recruitment and Costrecruit torepresent the cost of one recruit
Training The new members will take certain time in trainingbefore they are assigned to a cell and this action is denotedas Train119894 The time of training is denoted as 120590Training and thecost is Costtraining
Resource Acquirement The resources are acquired by themembers of resource cell We assume that the resourcein this model contains kinds of needs for performing atask The probability of resource acquirement is denoted as119875Resource and the cost is denoted as Costresource This action isrepresented as Res
The candidate action set of an agent is shown in Table 2
42 Task Workflow The purpose of terrorist organization isto produce terror events as much as possible and we simplifythe event to a task in this model The requirement of a taskis human and resource and they are simplified as agentsand resource The general workflow of terror organization ismodeled as the following four steps
Step 1 (generating task) The task is generated by the cellleader of action cell If an action cell is not operating anytask a new task will be created by the cell leader Eachtask 120593 requires amount of resource and agents whichcan be denoted as 120593(resource) and 120593(agent) respectivelyrepresenting the number of resources and agents required tofulfill a task
Step 2 (generating requirement) The requirement is evalu-ated by the cell leader and the element requirements will besent by the agents in the cell Assuming the size of action cell119866119894
AC is |119873119894AC| then the required number of agents is
119873Requirement = 120593 (agent) minus10038161003816100381610038161003816119873119894
AC10038161003816100381610038161003816 (1)
where119873Requirement is the number of generated human require-ments The number of required resources is 120593(resource) Theprocesses of human supplement and resource supplement aredifferent The human requirement will be directly sent to thesuperior and for the resource the leader will interact with allthemembers If there is a member connecting to the resource
cell the requirement will be sent to the member otherwiseit will be sent to the superior
Step 3 (agent response) After the requirements are sent theagent who receives the message will process the requirementbased on its role and relation Figure 3 shows the responsemechanism of an agent to a requirement The agent interactswith the other agent and it can get the information about theother agent In this response process the type of the link builtbetween two members in different cells is Relpp
Step 4 (task execution) Once the requirements are fulfilledthe action cell will launch an event and this is done by thecell leader of action cell The activity circle of action cell canbe illustrated as Figure 4 We hypothesize that the memberswill be lost in performing the task The probability for eachmember to be arrested is denoted as 119875arrest which dependson the efficiency of counterterror department
43 Organizational Recovery Mechanism The recovery pro-cess of terrorist network is the key mechanism for organiza-tion to survive Resilience is a dynamic process associatedwith systems that persist and perform their primary tasksunder pressure from exogenous shocks [32] We do notconsider the case of large-scale attack to the terrorist networkand focus on the normal consumption during the operationalprocess (a member can also be lost if he changes his mindwhich may happen because of the decay of his faith inthe organization In order to focus on the organizationaldynamics we hypothesize that the members will not be lostbecause of the belief issue) The loss in the task concernstwo types of cells the action cell and the resource cell Therecovery process is triggered after a task is performed Ourhypothesis is that the recovery mechanism comprises twobehaviors
(1) Check the leader of the action cell If the leader ofthe action cell is lost a new leader will be generatedfrom the remaining members The selection rule isbased on the attribute Age that is the agent whostays the longest in the cell will be designated as thenew leader This is because when an individual stayslonger in the organization it usually has the richestexperience (expertise knowledge etc) which is themost important characteristic of an action leader
(2) After that the new leader tries to build a new linkto the superior cell with the probability 119875reconnect Thetype of the newly built link is RelssThen the recoveryaction is finished
Figure 5 displays the overall recovery process of anaction cell It can be found that the recovery mechanismof the organization is partly finished by the organizationaloperational process
Resource cell also has recovery mechanism as the mem-ber may be lost in the process of performing a task In themodel we only consider the operational consumption andhypothesize that the member is only lost when the agent in
6 Discrete Dynamics in Nature and Society
If it is a member ofrequired cell
Find a neighborthat connects to the
required cell
Pass the requirement tothe agent
No
Yes
Pass the requirement tothe superior
No
If it can fulfil therequirement
Yes
Resource transferring
If it connects to therequiring cell
Yes
Yes
Find a member in thecell that can fulfilthe requirement
No
No
YesBuild a link to the select
agent in the requiring cell
No
Type of requirement
Resource
Human moving
Human
Receive a requirement
Finish a requirement
ActionChoice
Figure 3 The response mechanism of agent to a requirement
Generaterequirements
Fulfil therequirements Perform a task
Recovery Human loss
No loss
Figure 4 The activity circle of action cell
action cell who links to it is arrestedThe recoverymechanismof resource cell is triggered by the event that a member islost When a member is lost then a human requirement isgenerated by the cell leader
As illustrated in Figure 6 if member 1199096 who is in theaction cell and communicates with the resource sender 1199095 isarrested then 1199095 is lost tooThe leader of resource cell 1199091 then
sends human requirement to the conduct cell for supplementso as tomaintain functional completeness Even though somemembers are lost the resource cell can also acquire resourcebut the efficiency is lower than the complete cell
44 Organizational GrowthMechanism The launched eventslead to organizational growth and the correspondingincrease in size leads to faster production of new events[33] The more frequent the terrorist acts the faster thenetwork grows Events and social influence promote theappeal of organization and recruitment of new militants [12]For this consideration we assume that new action cell willbe generated after certain tasks are finished and a threshold120590Task is used as the trigger of the organizational growth Thesteps of growth can be modeled as follows
Step 1 (leader requirement) A human requirement is gener-ated and assigned to the member in the conduct cell who isin charge of action cell
Discrete Dynamics in Nature and Society 7
Lose members
Lose leader Generate new leader Human supplement
Human supplement New action cell
New action cell
Reconnection
Action cellPerform task
Perform task
Cell leaderx1
x1
x1
x1
x2
x2
x2
x2
x2
x3
x3x
3x3
x3
x3
x3
x3
x4
x5
x5
x5
x5
x5
x5 x
5
x5
x6
x6
x7
x7
x8
x8
x9
x9
Figure 5 The recovery process of action cell
Resourcesupplement
Action cell
Memberlost
Action cell
Resource cell Lose member Human supplement New resource cell
x1
x1x
1
x2
x2x
2
x3
x3x
3
x4
x4
x4
x5
x6
x1
x2
x3
x4
x5
x6
x7
x7
Figure 6 The recovery mechanism of resource cell
Step 2 (leader designation) After the target agent moving tothe new cell it is assigned as the leader of the new actioncell In this process the role of the agent is changed from amember of the training cell to the leader of the new cell
Step 3 (cell formation) Although the newly built action cellhas only one agent it can be performed as a cell whichmeansit can be assigned a task and generate the human and resourcerequirements and send it to the other agents The completeaction cell would form across the organizational task flow
The mechanism of organizational growth is shown inFigure 7 When the tasks performed by action cells reacha threshold 120590Task the network grows and establishes a newaction cell First the leader of the new cell (1199092) is assignedby the operational leader (1199091) in conduct cell Secondlyrequirement of human supplement is sent by the new leaderThirdly the members are transferred through the workflowof human supplement until the new action cell is formedSimultaneously the sizes of resource cell recruitment celland training cell all grow in order to satisfy the growingdemand of actions
5 Organizational Performance Evaluations
In order to analyze the organizational dynamics the indica-tors for evaluating performance are constructed based on thestructural model described in Section 3
As mentioned in [34] we evaluate the performance usingcost and benefit The cost consists of CostReP CostTrainingCostRecruit and CostResource These costs can be recordedduring the simulation and the total cost Costtotal can becalculated as
Costtotal = CostReP + CostTraining + CostRecruit
+ CostResource(2)
Organizational Size The most intuitional evaluation of thedevelopment of terrorist network is the organizational sizethat is the total number of agents in the terrorist organiza-tion and it can be calculated as follows
Size119866 = |119866|
=1003816100381610038161003816119866RC1003816100381610038161003816 +1003816100381610038161003816119866Co1003816100381610038161003816 +1003816100381610038161003816119866TC1003816100381610038161003816 +1003816100381610038161003816119866Re1003816100381610038161003816 + sum119894
10038161003816100381610038161003816119866119894
AC10038161003816100381610038161003816
(3)
Action CellsThe number of action cells measures the growthof terrorist organization from the operational aspectThis canbe measured through the value of 119899ac
Organizational Benefits The goal of terrorist organization isto perform attacks or take similar activities to reach politicalpurpose It is intuitional to use the number of completed tasksas the organizational benefits The task set Φ contains all the
8 Discrete Dynamics in Nature and Society
Conduct cell
Action leader
Conduct cell
Leader of newaction cell
Conduct cell Conduct cell
Conduct cell Assign leader Human supplement Generate new action cell
x1
x1x
1
x1
x2
x2
x2
x3
x3
x4
x4
x5
x6
Figure 7 The growth mechanism of terrorist network
tasks that have been performed The benefit can be obtainedusing
Benefit = |Φ| = 119896 (4)
Cost-Benefits Ratio This is to evaluate the cost-benefit of theactivity of performing tasks As the terrorist organizationalactivity is task driven we here only consider the requirementpassing behavior which measures the efficiency of how theterrorist is organizedThe cost-benefit is measured by the rateof cost and benefit which is denoted as CB so the cost-benefitof requirement passing CBReP is calculated as
CBReP =CostRePBenefit (5)
The other cost-benefits CBTraining CBRecruit andCBResource can also be calculated as (5)
6 Experimental Results and Discussions
In order to quantify the organizational dynamics we per-formed the experiments in a simulation environment anddiscussed the performance and the influence factors basedon the results Potential strategies for counterterrorism arediscussed with the results of sensitivity analysis
61 Experimental Setup The initial organizational structureof terrorist network was generated as shown in Figure 2 aswell as the roles of the individuals We tested the organiza-tional dynamics model in an open environment where thenumber of actions at each simulation step was not limitedIn order to carry out a comparative analysis we also useda limited environment with a parameter to represent thelimited cost 120590Cost which was the number of actions that canbe taken at each step in the network
(1) The initial terrorist network included a conduct cellthree action cells a resource cell a recruit cell and atraining cell The simulation step was 1000
(2) At each step cell leaders checked the situation of thecells and took actions based on the action set andworkflows (eg action cell performs the task if allthe requirements are finished) After that themessagelists of the agents who related to the actions wereupdated
(3) The structure of terrorist network was changed basedon the results of the cell actions (lose members linksetc)
(4) At each step the agents updated their message listsaccording to the interactive mechanism describedabove including message lists of themselves and themessage lists of the interactive objectsThe number ofactions which can be taken by an agent at one step isdenoted as 120590Behavior
(5) At each step the organization checks the number ofcompleted tasks If it reached a threshold 120590Task a newaction cell would be generated by the member in theconduct cell This means that a new action cell isgenerated at the time every 120590Task tasks are performed
(6) In the limited environments if the amount of actionsat each step reached a threshold 120590Cost the actions ofthe rest of agents at current step were stopped untilthe next step
For clarity the values of parameters that are used in thisexperiment and their explanations are listed in Table 3
62 Organizational Dynamics
621 Organizational Performance Figure 8 shows the sim-ulation results of organizational performance First theincrease of costs (Figure 8(a)) and tasks (Figure 8(b)) tendsto accelerate over time and this is because of the growth ofterrorist organization This is consistent with the results ofempirical studies that the violent events tend to acceleratewith increasing size and experience [33] There is a similarlinear relationship between the cost and completed tasksAlthough the cost during the processes of recruiting andresource acquirement depends on the probabilities whichare influenced by randomness the results show statisticallinear relationship It is easy to understand because theorganizational behaviors are driven by task Second thevalues of CostRecruit and CostResource are much higher thanCostTraining and this is caused by the probabilistic behav-iors of recruitment and resource acquirement The sum ofCostTraining CostRecruit and CostResource is much higher thanthat of CostReP and this is consistent with the results ofempirical studies that the preparation of a task is much moreexpensive than performing a task because multiple actionsare required to launch an attack Third Figure 8(a) showsthe differences caused by the task workflows CostRecruit andCostResource are influenced by the probabilities The action
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Discrete Dynamics in Nature and Society 3
Conductcell
Action cell
Recruit cellResource cell
Training cellHumansupplement
Humansupplement
Human supplement
Resourcesupplement
Action cellAction
cell
x1
x2
x3
x4
x5
x6
x7
x8
x9
x10
x11
x12
x13
x14
x14
x15
x16
Figure 2 Organizational structural model of terrorist network
activities in specific situations As the simulation time goeson the statistical results are changing and the organizationaldynamics analysis is carried out based on the results capturedin the system
3 The Organizational Structural Model ofTerrorist Network
Terrorist network is different from social network in twoaspects (a) cellular groups as basic activity units and (b)hierarchical command structure Terrorists operate accord-ing to this hybrid structure which makes the network do notcorrespond to a particular kind of networkmodel [25] Basedon the description of terrorist network in [26] a theoreticalterrorist network is shown in Figure 2The characteristics canbe understood from three aspects the cellular network struc-ture the hierarchical command structure and the individualactions based on the network
31 Cellular Network Structure It has been reported thatcellular network is the most suitable structure to describeterrorist organization [27 28] It usually consists of manycell groups (or subgroups) which are the basic active unitsto perform tasks [29 30] As shown in Figure 2 the cellsare cohesive structure of a set of connected individuals whocooperate to perform the task The reason for terrorists toadopt this network topology is structural compartmentaliza-tion [26] which is a trade-off between efficiency and security
Different network structures are adopted by the cellsto fulfil the diverse functional requirements [31] In thismodel we hypothesize that all the cellsrsquo structures conform tocorresponding network models Action cells have the struc-ture of fully connected network for frequent communicationbetween the members in it and so is the conduct cell Thestructures of recruit cells and resource cells conform to starnetwork which are helpful for the cells to perform task
(resource acquirement and human recruitment) widely andcovertlyWe also use star network here to depict the structureof training cell because of the function of the cell trainingthe terrorists singly as a ldquofactoryrdquo where the terrorists arethe products and are transferred to the other cells for humansupplement
The terrorist network can be regarded as a graph 119866 =(119873 119864) where 119873 represents the set of agents (terrorists)and 119864 is the set of all the relations in the network Wehypothesize that the network only includes the member ofthe organization whichmeans the agents in the environmentare not considered Only the formal relations are included inthis model and the informal relations such as social relationare not considered in this paper The cells can be regarded assubgroups of119866 In thismodel five types of cell are consideredaccording to their function in the organization and thenotations are listed inTable 1 aswell as the explanations Eachcell contains a node set (as 119873AC in 119866AC) and an edge set (as119864AC in 119866AC) which is the set of all the relations between thenodes in the cell As there is more than one action cell we use119899ac to represent the number of action cells
32 Hierarchical Command Structure
321 Role of Terrorist and Type of Relation Recall that terror-ist network is different from social network in organizationalactivities As an organization the roles of terrorists are diverseand the relational type depends on their roles Based on thecellular network structure we define two types of roles in themodel the cell leader and the cell memberThe role is definedas a property of agent and the definitions are as follows
Cell leader (denoted as Role119897) is the leader of a cell Itorganizes the activities of the cell members and connects tothe superior cell for organizational command
Cell member (denoted as Role119898) is the member of a cellIt performs the specific activities in task process
We also define two types of relations in this model thepeer-to-peer relation and the superior-subordinate relation
Peer-to-peer relation (denoted as Relpp) is the type of linkbetween the agents with the same role
Superior-subordinate relation (denoted as Relss) is thetype of link between the agents with different roles
As the terrorist organization has hierarchical structurewe give the assumptions used in this model
(1) The type of the relation between the members in thesame cell is Relpp and the type of the relation betweenthem and the cell leader is Relss
(2) The member of the conduct cell is the superior of theother cells which means the relations between themand the leaders of the other cells are Relss
(3) There is no relation between the cells leaders that isa cell leader cannot interact directly with another cellleader
These assumptions are given based on the characteristicsof terrorist organization and empirical research According tothe definition and described cellular network in Section 31the hierarchical structure of terrorist network is fixed to some
4 Discrete Dynamics in Nature and Society
Table 1 The notations of the cells in the cellular network
Notation Explanations119866AC Action cell and 119866AC = (119873AC 119864AC) sub 119866119866RC Resource cell and 119866RC = (119873RC 119864RC) sub 119866119866Co Conduct cell and 119866Co = (119873Co 119864Co) sub 119866119866TC Training cell and 119866TC = (119873TC 119864TC) sub 119866119866Re Recruit cell and 119866Re = (119873Re 119864Re) sub 119866119866119894
AC The 119894th action cell and 119866119894AC = (119873119894
AC 119864119894
AC) sub 119866
extent The conduct cell is barely changed and the roles ofthe members are specified The leaders in the other cells aredesignated by the superior in the conduct cell which is tosome extent fixed as well
322 Intercell Hierarchical Structure The command relationbetween cells represents inherently hierarchies between func-tional compartments As illustrated in Figure 2 the tasks areoperated by the conduct cell with the instructions sent tothe other cells Firstly the conduct cell sends missions tothe action cells that then take actions required to performthe tasks When the conduct cell receives the feedback ofhuman and resource requirements it passes human andresource supplements to the training cell and resource cellseparately This hierarchical command structure makes thecore of organization able to coordinate the members for anoverall goal and maintain their own agendas Note that theorders are transmitted by the individuals in cells because thecell is not a real entity but an organizational form of groupmembers
323 Intracell Hierarchical Structure Decentralization attactical level is due to the difficulty of real-time commandand control within a large clandestine cellular network [26]Cell leaders usually have freedom of tactical decision-makingand action based on local condition which produces intracellhierarchies and self-organization As shown in Figure 2leaders of functional cells (1199095 1199096 11990910) are receivers ofthe assigned missions They organize the actions of othermembers in cells to complete tasks Take action cell (where 1199095stays) as an example 1199095 receives a mission from the superiorcell (conduct cell) then it evaluates the requirements basedon the condition of the action cell and sends the feedback tothe superior cell During the action period 11990911 is in charge ofreceiving resource from the resource cell and acts under thecommand of 1199095 because it is the only one that communicateswith the superior cell Even in the core cell the members(1199092 1199093 1199094) are under the command of the cell leader (1199091) whois either the leader of the organization or the only one thatconnects to the superior leader
4 The Organizational Process Model ofTerrorist Network
The organizational process model is proposed to describe theoperational mechanism of terrorist network First the agentmodel is given and the actions are defined as a set of available
actions Then the response of agent in different mechanismis givenThe organizational processmodel can be understoodas the interactional rule of agent [17]
41 The Agent Model The system considers a populationof agents (terrorists) that continually act as the model pro-gresses As a terrorist organization the individual behaviorsand attributes are complicated For clarity we only considerthe following attributes which are related to the organiza-tional dynamics in this model
Role This is the role of agent in the cell At the beginningthe leader of a cell is assigned and can be identified by theindividual structure Once the leader of a cell is lost a newleader will be generated in the cell and this will be discussedin the recovery mechanism
Age This is the length of time that an agent stays in thenetwork
Resource This is the attribute which re-presents if an agentholds resource
Cell Type This is the type of the cell which the agent belongsto
The actions are limited by the role and the cell typeof agent which means that agents have different candidateaction set according to their position in the organizationTheactions used in the model are defined as follows
Requirement PassingThe requirement for resource or humanis passed from the source agent to the target agent Thisbehavior is defined as ReP119894119895 which represents that therequirement is passed from agent 119894 to agent 119895 The cost ofpassing a requirement is CostReP
Resource Transferring The required resource is transferredby terrorists through established channels (solid arrows inFigure 2) The element transferring behavior is defined asTra119894119895 where 119894 and 119895 are the terrorists transferring the element
Moving The agent moves from one cell to another cellfor human supplement This behavior is denoted as Mo119894and 119894 is the agent There are some differences between theresource transferring and the moving Agent remains in theresource cell and it links to the required cell after resource istransferred Once an agent moves to a new cell it would notconnect to the original cell whichmeans the other cell cannotdirectly send the human requirement to the training cell butit can send the resource requirement to the resource cell aslong as the link exists
Building a Link This action is taken when a link is neededbetween two agents and it is denoted as BuL119894119895 where 119894 and119895 are the agents who built the relation As there are two typesof links the newly built link depends on the situation andthe roles of the two agents which will be discussed in thefollowing organizational mechanism
Discrete Dynamics in Nature and Society 5
Table 2 The candidate action set of an agent in terrorist network
Role119897
Role119898
Action cell ReP BuL ReP Tra BuLResource cell ReP BuL Res ReP Tra BuL ResConduct cell ReP ReP BuLTraining cell ReP BuL Mo BuL Train RePRecruit cell ReP BuL Rec Mo BuL Rec ReP
Recruitment Recruitment (denoted as Rec) is performed bythe members of recruit cell and it is the only way to generatenew member of terrorist network We use 119875Recruit to denotethe probability of successful recruitment and Costrecruit torepresent the cost of one recruit
Training The new members will take certain time in trainingbefore they are assigned to a cell and this action is denotedas Train119894 The time of training is denoted as 120590Training and thecost is Costtraining
Resource Acquirement The resources are acquired by themembers of resource cell We assume that the resourcein this model contains kinds of needs for performing atask The probability of resource acquirement is denoted as119875Resource and the cost is denoted as Costresource This action isrepresented as Res
The candidate action set of an agent is shown in Table 2
42 Task Workflow The purpose of terrorist organization isto produce terror events as much as possible and we simplifythe event to a task in this model The requirement of a taskis human and resource and they are simplified as agentsand resource The general workflow of terror organization ismodeled as the following four steps
Step 1 (generating task) The task is generated by the cellleader of action cell If an action cell is not operating anytask a new task will be created by the cell leader Eachtask 120593 requires amount of resource and agents whichcan be denoted as 120593(resource) and 120593(agent) respectivelyrepresenting the number of resources and agents required tofulfill a task
Step 2 (generating requirement) The requirement is evalu-ated by the cell leader and the element requirements will besent by the agents in the cell Assuming the size of action cell119866119894
AC is |119873119894AC| then the required number of agents is
119873Requirement = 120593 (agent) minus10038161003816100381610038161003816119873119894
AC10038161003816100381610038161003816 (1)
where119873Requirement is the number of generated human require-ments The number of required resources is 120593(resource) Theprocesses of human supplement and resource supplement aredifferent The human requirement will be directly sent to thesuperior and for the resource the leader will interact with allthemembers If there is a member connecting to the resource
cell the requirement will be sent to the member otherwiseit will be sent to the superior
Step 3 (agent response) After the requirements are sent theagent who receives the message will process the requirementbased on its role and relation Figure 3 shows the responsemechanism of an agent to a requirement The agent interactswith the other agent and it can get the information about theother agent In this response process the type of the link builtbetween two members in different cells is Relpp
Step 4 (task execution) Once the requirements are fulfilledthe action cell will launch an event and this is done by thecell leader of action cell The activity circle of action cell canbe illustrated as Figure 4 We hypothesize that the memberswill be lost in performing the task The probability for eachmember to be arrested is denoted as 119875arrest which dependson the efficiency of counterterror department
43 Organizational Recovery Mechanism The recovery pro-cess of terrorist network is the key mechanism for organiza-tion to survive Resilience is a dynamic process associatedwith systems that persist and perform their primary tasksunder pressure from exogenous shocks [32] We do notconsider the case of large-scale attack to the terrorist networkand focus on the normal consumption during the operationalprocess (a member can also be lost if he changes his mindwhich may happen because of the decay of his faith inthe organization In order to focus on the organizationaldynamics we hypothesize that the members will not be lostbecause of the belief issue) The loss in the task concernstwo types of cells the action cell and the resource cell Therecovery process is triggered after a task is performed Ourhypothesis is that the recovery mechanism comprises twobehaviors
(1) Check the leader of the action cell If the leader ofthe action cell is lost a new leader will be generatedfrom the remaining members The selection rule isbased on the attribute Age that is the agent whostays the longest in the cell will be designated as thenew leader This is because when an individual stayslonger in the organization it usually has the richestexperience (expertise knowledge etc) which is themost important characteristic of an action leader
(2) After that the new leader tries to build a new linkto the superior cell with the probability 119875reconnect Thetype of the newly built link is RelssThen the recoveryaction is finished
Figure 5 displays the overall recovery process of anaction cell It can be found that the recovery mechanismof the organization is partly finished by the organizationaloperational process
Resource cell also has recovery mechanism as the mem-ber may be lost in the process of performing a task In themodel we only consider the operational consumption andhypothesize that the member is only lost when the agent in
6 Discrete Dynamics in Nature and Society
If it is a member ofrequired cell
Find a neighborthat connects to the
required cell
Pass the requirement tothe agent
No
Yes
Pass the requirement tothe superior
No
If it can fulfil therequirement
Yes
Resource transferring
If it connects to therequiring cell
Yes
Yes
Find a member in thecell that can fulfilthe requirement
No
No
YesBuild a link to the select
agent in the requiring cell
No
Type of requirement
Resource
Human moving
Human
Receive a requirement
Finish a requirement
ActionChoice
Figure 3 The response mechanism of agent to a requirement
Generaterequirements
Fulfil therequirements Perform a task
Recovery Human loss
No loss
Figure 4 The activity circle of action cell
action cell who links to it is arrestedThe recoverymechanismof resource cell is triggered by the event that a member islost When a member is lost then a human requirement isgenerated by the cell leader
As illustrated in Figure 6 if member 1199096 who is in theaction cell and communicates with the resource sender 1199095 isarrested then 1199095 is lost tooThe leader of resource cell 1199091 then
sends human requirement to the conduct cell for supplementso as tomaintain functional completeness Even though somemembers are lost the resource cell can also acquire resourcebut the efficiency is lower than the complete cell
44 Organizational GrowthMechanism The launched eventslead to organizational growth and the correspondingincrease in size leads to faster production of new events[33] The more frequent the terrorist acts the faster thenetwork grows Events and social influence promote theappeal of organization and recruitment of new militants [12]For this consideration we assume that new action cell willbe generated after certain tasks are finished and a threshold120590Task is used as the trigger of the organizational growth Thesteps of growth can be modeled as follows
Step 1 (leader requirement) A human requirement is gener-ated and assigned to the member in the conduct cell who isin charge of action cell
Discrete Dynamics in Nature and Society 7
Lose members
Lose leader Generate new leader Human supplement
Human supplement New action cell
New action cell
Reconnection
Action cellPerform task
Perform task
Cell leaderx1
x1
x1
x1
x2
x2
x2
x2
x2
x3
x3x
3x3
x3
x3
x3
x3
x4
x5
x5
x5
x5
x5
x5 x
5
x5
x6
x6
x7
x7
x8
x8
x9
x9
Figure 5 The recovery process of action cell
Resourcesupplement
Action cell
Memberlost
Action cell
Resource cell Lose member Human supplement New resource cell
x1
x1x
1
x2
x2x
2
x3
x3x
3
x4
x4
x4
x5
x6
x1
x2
x3
x4
x5
x6
x7
x7
Figure 6 The recovery mechanism of resource cell
Step 2 (leader designation) After the target agent moving tothe new cell it is assigned as the leader of the new actioncell In this process the role of the agent is changed from amember of the training cell to the leader of the new cell
Step 3 (cell formation) Although the newly built action cellhas only one agent it can be performed as a cell whichmeansit can be assigned a task and generate the human and resourcerequirements and send it to the other agents The completeaction cell would form across the organizational task flow
The mechanism of organizational growth is shown inFigure 7 When the tasks performed by action cells reacha threshold 120590Task the network grows and establishes a newaction cell First the leader of the new cell (1199092) is assignedby the operational leader (1199091) in conduct cell Secondlyrequirement of human supplement is sent by the new leaderThirdly the members are transferred through the workflowof human supplement until the new action cell is formedSimultaneously the sizes of resource cell recruitment celland training cell all grow in order to satisfy the growingdemand of actions
5 Organizational Performance Evaluations
In order to analyze the organizational dynamics the indica-tors for evaluating performance are constructed based on thestructural model described in Section 3
As mentioned in [34] we evaluate the performance usingcost and benefit The cost consists of CostReP CostTrainingCostRecruit and CostResource These costs can be recordedduring the simulation and the total cost Costtotal can becalculated as
Costtotal = CostReP + CostTraining + CostRecruit
+ CostResource(2)
Organizational Size The most intuitional evaluation of thedevelopment of terrorist network is the organizational sizethat is the total number of agents in the terrorist organiza-tion and it can be calculated as follows
Size119866 = |119866|
=1003816100381610038161003816119866RC1003816100381610038161003816 +1003816100381610038161003816119866Co1003816100381610038161003816 +1003816100381610038161003816119866TC1003816100381610038161003816 +1003816100381610038161003816119866Re1003816100381610038161003816 + sum119894
10038161003816100381610038161003816119866119894
AC10038161003816100381610038161003816
(3)
Action CellsThe number of action cells measures the growthof terrorist organization from the operational aspectThis canbe measured through the value of 119899ac
Organizational Benefits The goal of terrorist organization isto perform attacks or take similar activities to reach politicalpurpose It is intuitional to use the number of completed tasksas the organizational benefits The task set Φ contains all the
8 Discrete Dynamics in Nature and Society
Conduct cell
Action leader
Conduct cell
Leader of newaction cell
Conduct cell Conduct cell
Conduct cell Assign leader Human supplement Generate new action cell
x1
x1x
1
x1
x2
x2
x2
x3
x3
x4
x4
x5
x6
Figure 7 The growth mechanism of terrorist network
tasks that have been performed The benefit can be obtainedusing
Benefit = |Φ| = 119896 (4)
Cost-Benefits Ratio This is to evaluate the cost-benefit of theactivity of performing tasks As the terrorist organizationalactivity is task driven we here only consider the requirementpassing behavior which measures the efficiency of how theterrorist is organizedThe cost-benefit is measured by the rateof cost and benefit which is denoted as CB so the cost-benefitof requirement passing CBReP is calculated as
CBReP =CostRePBenefit (5)
The other cost-benefits CBTraining CBRecruit andCBResource can also be calculated as (5)
6 Experimental Results and Discussions
In order to quantify the organizational dynamics we per-formed the experiments in a simulation environment anddiscussed the performance and the influence factors basedon the results Potential strategies for counterterrorism arediscussed with the results of sensitivity analysis
61 Experimental Setup The initial organizational structureof terrorist network was generated as shown in Figure 2 aswell as the roles of the individuals We tested the organiza-tional dynamics model in an open environment where thenumber of actions at each simulation step was not limitedIn order to carry out a comparative analysis we also useda limited environment with a parameter to represent thelimited cost 120590Cost which was the number of actions that canbe taken at each step in the network
(1) The initial terrorist network included a conduct cellthree action cells a resource cell a recruit cell and atraining cell The simulation step was 1000
(2) At each step cell leaders checked the situation of thecells and took actions based on the action set andworkflows (eg action cell performs the task if allthe requirements are finished) After that themessagelists of the agents who related to the actions wereupdated
(3) The structure of terrorist network was changed basedon the results of the cell actions (lose members linksetc)
(4) At each step the agents updated their message listsaccording to the interactive mechanism describedabove including message lists of themselves and themessage lists of the interactive objectsThe number ofactions which can be taken by an agent at one step isdenoted as 120590Behavior
(5) At each step the organization checks the number ofcompleted tasks If it reached a threshold 120590Task a newaction cell would be generated by the member in theconduct cell This means that a new action cell isgenerated at the time every 120590Task tasks are performed
(6) In the limited environments if the amount of actionsat each step reached a threshold 120590Cost the actions ofthe rest of agents at current step were stopped untilthe next step
For clarity the values of parameters that are used in thisexperiment and their explanations are listed in Table 3
62 Organizational Dynamics
621 Organizational Performance Figure 8 shows the sim-ulation results of organizational performance First theincrease of costs (Figure 8(a)) and tasks (Figure 8(b)) tendsto accelerate over time and this is because of the growth ofterrorist organization This is consistent with the results ofempirical studies that the violent events tend to acceleratewith increasing size and experience [33] There is a similarlinear relationship between the cost and completed tasksAlthough the cost during the processes of recruiting andresource acquirement depends on the probabilities whichare influenced by randomness the results show statisticallinear relationship It is easy to understand because theorganizational behaviors are driven by task Second thevalues of CostRecruit and CostResource are much higher thanCostTraining and this is caused by the probabilistic behav-iors of recruitment and resource acquirement The sum ofCostTraining CostRecruit and CostResource is much higher thanthat of CostReP and this is consistent with the results ofempirical studies that the preparation of a task is much moreexpensive than performing a task because multiple actionsare required to launch an attack Third Figure 8(a) showsthe differences caused by the task workflows CostRecruit andCostResource are influenced by the probabilities The action
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
4 Discrete Dynamics in Nature and Society
Table 1 The notations of the cells in the cellular network
Notation Explanations119866AC Action cell and 119866AC = (119873AC 119864AC) sub 119866119866RC Resource cell and 119866RC = (119873RC 119864RC) sub 119866119866Co Conduct cell and 119866Co = (119873Co 119864Co) sub 119866119866TC Training cell and 119866TC = (119873TC 119864TC) sub 119866119866Re Recruit cell and 119866Re = (119873Re 119864Re) sub 119866119866119894
AC The 119894th action cell and 119866119894AC = (119873119894
AC 119864119894
AC) sub 119866
extent The conduct cell is barely changed and the roles ofthe members are specified The leaders in the other cells aredesignated by the superior in the conduct cell which is tosome extent fixed as well
322 Intercell Hierarchical Structure The command relationbetween cells represents inherently hierarchies between func-tional compartments As illustrated in Figure 2 the tasks areoperated by the conduct cell with the instructions sent tothe other cells Firstly the conduct cell sends missions tothe action cells that then take actions required to performthe tasks When the conduct cell receives the feedback ofhuman and resource requirements it passes human andresource supplements to the training cell and resource cellseparately This hierarchical command structure makes thecore of organization able to coordinate the members for anoverall goal and maintain their own agendas Note that theorders are transmitted by the individuals in cells because thecell is not a real entity but an organizational form of groupmembers
323 Intracell Hierarchical Structure Decentralization attactical level is due to the difficulty of real-time commandand control within a large clandestine cellular network [26]Cell leaders usually have freedom of tactical decision-makingand action based on local condition which produces intracellhierarchies and self-organization As shown in Figure 2leaders of functional cells (1199095 1199096 11990910) are receivers ofthe assigned missions They organize the actions of othermembers in cells to complete tasks Take action cell (where 1199095stays) as an example 1199095 receives a mission from the superiorcell (conduct cell) then it evaluates the requirements basedon the condition of the action cell and sends the feedback tothe superior cell During the action period 11990911 is in charge ofreceiving resource from the resource cell and acts under thecommand of 1199095 because it is the only one that communicateswith the superior cell Even in the core cell the members(1199092 1199093 1199094) are under the command of the cell leader (1199091) whois either the leader of the organization or the only one thatconnects to the superior leader
4 The Organizational Process Model ofTerrorist Network
The organizational process model is proposed to describe theoperational mechanism of terrorist network First the agentmodel is given and the actions are defined as a set of available
actions Then the response of agent in different mechanismis givenThe organizational processmodel can be understoodas the interactional rule of agent [17]
41 The Agent Model The system considers a populationof agents (terrorists) that continually act as the model pro-gresses As a terrorist organization the individual behaviorsand attributes are complicated For clarity we only considerthe following attributes which are related to the organiza-tional dynamics in this model
Role This is the role of agent in the cell At the beginningthe leader of a cell is assigned and can be identified by theindividual structure Once the leader of a cell is lost a newleader will be generated in the cell and this will be discussedin the recovery mechanism
Age This is the length of time that an agent stays in thenetwork
Resource This is the attribute which re-presents if an agentholds resource
Cell Type This is the type of the cell which the agent belongsto
The actions are limited by the role and the cell typeof agent which means that agents have different candidateaction set according to their position in the organizationTheactions used in the model are defined as follows
Requirement PassingThe requirement for resource or humanis passed from the source agent to the target agent Thisbehavior is defined as ReP119894119895 which represents that therequirement is passed from agent 119894 to agent 119895 The cost ofpassing a requirement is CostReP
Resource Transferring The required resource is transferredby terrorists through established channels (solid arrows inFigure 2) The element transferring behavior is defined asTra119894119895 where 119894 and 119895 are the terrorists transferring the element
Moving The agent moves from one cell to another cellfor human supplement This behavior is denoted as Mo119894and 119894 is the agent There are some differences between theresource transferring and the moving Agent remains in theresource cell and it links to the required cell after resource istransferred Once an agent moves to a new cell it would notconnect to the original cell whichmeans the other cell cannotdirectly send the human requirement to the training cell butit can send the resource requirement to the resource cell aslong as the link exists
Building a Link This action is taken when a link is neededbetween two agents and it is denoted as BuL119894119895 where 119894 and119895 are the agents who built the relation As there are two typesof links the newly built link depends on the situation andthe roles of the two agents which will be discussed in thefollowing organizational mechanism
Discrete Dynamics in Nature and Society 5
Table 2 The candidate action set of an agent in terrorist network
Role119897
Role119898
Action cell ReP BuL ReP Tra BuLResource cell ReP BuL Res ReP Tra BuL ResConduct cell ReP ReP BuLTraining cell ReP BuL Mo BuL Train RePRecruit cell ReP BuL Rec Mo BuL Rec ReP
Recruitment Recruitment (denoted as Rec) is performed bythe members of recruit cell and it is the only way to generatenew member of terrorist network We use 119875Recruit to denotethe probability of successful recruitment and Costrecruit torepresent the cost of one recruit
Training The new members will take certain time in trainingbefore they are assigned to a cell and this action is denotedas Train119894 The time of training is denoted as 120590Training and thecost is Costtraining
Resource Acquirement The resources are acquired by themembers of resource cell We assume that the resourcein this model contains kinds of needs for performing atask The probability of resource acquirement is denoted as119875Resource and the cost is denoted as Costresource This action isrepresented as Res
The candidate action set of an agent is shown in Table 2
42 Task Workflow The purpose of terrorist organization isto produce terror events as much as possible and we simplifythe event to a task in this model The requirement of a taskis human and resource and they are simplified as agentsand resource The general workflow of terror organization ismodeled as the following four steps
Step 1 (generating task) The task is generated by the cellleader of action cell If an action cell is not operating anytask a new task will be created by the cell leader Eachtask 120593 requires amount of resource and agents whichcan be denoted as 120593(resource) and 120593(agent) respectivelyrepresenting the number of resources and agents required tofulfill a task
Step 2 (generating requirement) The requirement is evalu-ated by the cell leader and the element requirements will besent by the agents in the cell Assuming the size of action cell119866119894
AC is |119873119894AC| then the required number of agents is
119873Requirement = 120593 (agent) minus10038161003816100381610038161003816119873119894
AC10038161003816100381610038161003816 (1)
where119873Requirement is the number of generated human require-ments The number of required resources is 120593(resource) Theprocesses of human supplement and resource supplement aredifferent The human requirement will be directly sent to thesuperior and for the resource the leader will interact with allthemembers If there is a member connecting to the resource
cell the requirement will be sent to the member otherwiseit will be sent to the superior
Step 3 (agent response) After the requirements are sent theagent who receives the message will process the requirementbased on its role and relation Figure 3 shows the responsemechanism of an agent to a requirement The agent interactswith the other agent and it can get the information about theother agent In this response process the type of the link builtbetween two members in different cells is Relpp
Step 4 (task execution) Once the requirements are fulfilledthe action cell will launch an event and this is done by thecell leader of action cell The activity circle of action cell canbe illustrated as Figure 4 We hypothesize that the memberswill be lost in performing the task The probability for eachmember to be arrested is denoted as 119875arrest which dependson the efficiency of counterterror department
43 Organizational Recovery Mechanism The recovery pro-cess of terrorist network is the key mechanism for organiza-tion to survive Resilience is a dynamic process associatedwith systems that persist and perform their primary tasksunder pressure from exogenous shocks [32] We do notconsider the case of large-scale attack to the terrorist networkand focus on the normal consumption during the operationalprocess (a member can also be lost if he changes his mindwhich may happen because of the decay of his faith inthe organization In order to focus on the organizationaldynamics we hypothesize that the members will not be lostbecause of the belief issue) The loss in the task concernstwo types of cells the action cell and the resource cell Therecovery process is triggered after a task is performed Ourhypothesis is that the recovery mechanism comprises twobehaviors
(1) Check the leader of the action cell If the leader ofthe action cell is lost a new leader will be generatedfrom the remaining members The selection rule isbased on the attribute Age that is the agent whostays the longest in the cell will be designated as thenew leader This is because when an individual stayslonger in the organization it usually has the richestexperience (expertise knowledge etc) which is themost important characteristic of an action leader
(2) After that the new leader tries to build a new linkto the superior cell with the probability 119875reconnect Thetype of the newly built link is RelssThen the recoveryaction is finished
Figure 5 displays the overall recovery process of anaction cell It can be found that the recovery mechanismof the organization is partly finished by the organizationaloperational process
Resource cell also has recovery mechanism as the mem-ber may be lost in the process of performing a task In themodel we only consider the operational consumption andhypothesize that the member is only lost when the agent in
6 Discrete Dynamics in Nature and Society
If it is a member ofrequired cell
Find a neighborthat connects to the
required cell
Pass the requirement tothe agent
No
Yes
Pass the requirement tothe superior
No
If it can fulfil therequirement
Yes
Resource transferring
If it connects to therequiring cell
Yes
Yes
Find a member in thecell that can fulfilthe requirement
No
No
YesBuild a link to the select
agent in the requiring cell
No
Type of requirement
Resource
Human moving
Human
Receive a requirement
Finish a requirement
ActionChoice
Figure 3 The response mechanism of agent to a requirement
Generaterequirements
Fulfil therequirements Perform a task
Recovery Human loss
No loss
Figure 4 The activity circle of action cell
action cell who links to it is arrestedThe recoverymechanismof resource cell is triggered by the event that a member islost When a member is lost then a human requirement isgenerated by the cell leader
As illustrated in Figure 6 if member 1199096 who is in theaction cell and communicates with the resource sender 1199095 isarrested then 1199095 is lost tooThe leader of resource cell 1199091 then
sends human requirement to the conduct cell for supplementso as tomaintain functional completeness Even though somemembers are lost the resource cell can also acquire resourcebut the efficiency is lower than the complete cell
44 Organizational GrowthMechanism The launched eventslead to organizational growth and the correspondingincrease in size leads to faster production of new events[33] The more frequent the terrorist acts the faster thenetwork grows Events and social influence promote theappeal of organization and recruitment of new militants [12]For this consideration we assume that new action cell willbe generated after certain tasks are finished and a threshold120590Task is used as the trigger of the organizational growth Thesteps of growth can be modeled as follows
Step 1 (leader requirement) A human requirement is gener-ated and assigned to the member in the conduct cell who isin charge of action cell
Discrete Dynamics in Nature and Society 7
Lose members
Lose leader Generate new leader Human supplement
Human supplement New action cell
New action cell
Reconnection
Action cellPerform task
Perform task
Cell leaderx1
x1
x1
x1
x2
x2
x2
x2
x2
x3
x3x
3x3
x3
x3
x3
x3
x4
x5
x5
x5
x5
x5
x5 x
5
x5
x6
x6
x7
x7
x8
x8
x9
x9
Figure 5 The recovery process of action cell
Resourcesupplement
Action cell
Memberlost
Action cell
Resource cell Lose member Human supplement New resource cell
x1
x1x
1
x2
x2x
2
x3
x3x
3
x4
x4
x4
x5
x6
x1
x2
x3
x4
x5
x6
x7
x7
Figure 6 The recovery mechanism of resource cell
Step 2 (leader designation) After the target agent moving tothe new cell it is assigned as the leader of the new actioncell In this process the role of the agent is changed from amember of the training cell to the leader of the new cell
Step 3 (cell formation) Although the newly built action cellhas only one agent it can be performed as a cell whichmeansit can be assigned a task and generate the human and resourcerequirements and send it to the other agents The completeaction cell would form across the organizational task flow
The mechanism of organizational growth is shown inFigure 7 When the tasks performed by action cells reacha threshold 120590Task the network grows and establishes a newaction cell First the leader of the new cell (1199092) is assignedby the operational leader (1199091) in conduct cell Secondlyrequirement of human supplement is sent by the new leaderThirdly the members are transferred through the workflowof human supplement until the new action cell is formedSimultaneously the sizes of resource cell recruitment celland training cell all grow in order to satisfy the growingdemand of actions
5 Organizational Performance Evaluations
In order to analyze the organizational dynamics the indica-tors for evaluating performance are constructed based on thestructural model described in Section 3
As mentioned in [34] we evaluate the performance usingcost and benefit The cost consists of CostReP CostTrainingCostRecruit and CostResource These costs can be recordedduring the simulation and the total cost Costtotal can becalculated as
Costtotal = CostReP + CostTraining + CostRecruit
+ CostResource(2)
Organizational Size The most intuitional evaluation of thedevelopment of terrorist network is the organizational sizethat is the total number of agents in the terrorist organiza-tion and it can be calculated as follows
Size119866 = |119866|
=1003816100381610038161003816119866RC1003816100381610038161003816 +1003816100381610038161003816119866Co1003816100381610038161003816 +1003816100381610038161003816119866TC1003816100381610038161003816 +1003816100381610038161003816119866Re1003816100381610038161003816 + sum119894
10038161003816100381610038161003816119866119894
AC10038161003816100381610038161003816
(3)
Action CellsThe number of action cells measures the growthof terrorist organization from the operational aspectThis canbe measured through the value of 119899ac
Organizational Benefits The goal of terrorist organization isto perform attacks or take similar activities to reach politicalpurpose It is intuitional to use the number of completed tasksas the organizational benefits The task set Φ contains all the
8 Discrete Dynamics in Nature and Society
Conduct cell
Action leader
Conduct cell
Leader of newaction cell
Conduct cell Conduct cell
Conduct cell Assign leader Human supplement Generate new action cell
x1
x1x
1
x1
x2
x2
x2
x3
x3
x4
x4
x5
x6
Figure 7 The growth mechanism of terrorist network
tasks that have been performed The benefit can be obtainedusing
Benefit = |Φ| = 119896 (4)
Cost-Benefits Ratio This is to evaluate the cost-benefit of theactivity of performing tasks As the terrorist organizationalactivity is task driven we here only consider the requirementpassing behavior which measures the efficiency of how theterrorist is organizedThe cost-benefit is measured by the rateof cost and benefit which is denoted as CB so the cost-benefitof requirement passing CBReP is calculated as
CBReP =CostRePBenefit (5)
The other cost-benefits CBTraining CBRecruit andCBResource can also be calculated as (5)
6 Experimental Results and Discussions
In order to quantify the organizational dynamics we per-formed the experiments in a simulation environment anddiscussed the performance and the influence factors basedon the results Potential strategies for counterterrorism arediscussed with the results of sensitivity analysis
61 Experimental Setup The initial organizational structureof terrorist network was generated as shown in Figure 2 aswell as the roles of the individuals We tested the organiza-tional dynamics model in an open environment where thenumber of actions at each simulation step was not limitedIn order to carry out a comparative analysis we also useda limited environment with a parameter to represent thelimited cost 120590Cost which was the number of actions that canbe taken at each step in the network
(1) The initial terrorist network included a conduct cellthree action cells a resource cell a recruit cell and atraining cell The simulation step was 1000
(2) At each step cell leaders checked the situation of thecells and took actions based on the action set andworkflows (eg action cell performs the task if allthe requirements are finished) After that themessagelists of the agents who related to the actions wereupdated
(3) The structure of terrorist network was changed basedon the results of the cell actions (lose members linksetc)
(4) At each step the agents updated their message listsaccording to the interactive mechanism describedabove including message lists of themselves and themessage lists of the interactive objectsThe number ofactions which can be taken by an agent at one step isdenoted as 120590Behavior
(5) At each step the organization checks the number ofcompleted tasks If it reached a threshold 120590Task a newaction cell would be generated by the member in theconduct cell This means that a new action cell isgenerated at the time every 120590Task tasks are performed
(6) In the limited environments if the amount of actionsat each step reached a threshold 120590Cost the actions ofthe rest of agents at current step were stopped untilthe next step
For clarity the values of parameters that are used in thisexperiment and their explanations are listed in Table 3
62 Organizational Dynamics
621 Organizational Performance Figure 8 shows the sim-ulation results of organizational performance First theincrease of costs (Figure 8(a)) and tasks (Figure 8(b)) tendsto accelerate over time and this is because of the growth ofterrorist organization This is consistent with the results ofempirical studies that the violent events tend to acceleratewith increasing size and experience [33] There is a similarlinear relationship between the cost and completed tasksAlthough the cost during the processes of recruiting andresource acquirement depends on the probabilities whichare influenced by randomness the results show statisticallinear relationship It is easy to understand because theorganizational behaviors are driven by task Second thevalues of CostRecruit and CostResource are much higher thanCostTraining and this is caused by the probabilistic behav-iors of recruitment and resource acquirement The sum ofCostTraining CostRecruit and CostResource is much higher thanthat of CostReP and this is consistent with the results ofempirical studies that the preparation of a task is much moreexpensive than performing a task because multiple actionsare required to launch an attack Third Figure 8(a) showsthe differences caused by the task workflows CostRecruit andCostResource are influenced by the probabilities The action
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Discrete Dynamics in Nature and Society 5
Table 2 The candidate action set of an agent in terrorist network
Role119897
Role119898
Action cell ReP BuL ReP Tra BuLResource cell ReP BuL Res ReP Tra BuL ResConduct cell ReP ReP BuLTraining cell ReP BuL Mo BuL Train RePRecruit cell ReP BuL Rec Mo BuL Rec ReP
Recruitment Recruitment (denoted as Rec) is performed bythe members of recruit cell and it is the only way to generatenew member of terrorist network We use 119875Recruit to denotethe probability of successful recruitment and Costrecruit torepresent the cost of one recruit
Training The new members will take certain time in trainingbefore they are assigned to a cell and this action is denotedas Train119894 The time of training is denoted as 120590Training and thecost is Costtraining
Resource Acquirement The resources are acquired by themembers of resource cell We assume that the resourcein this model contains kinds of needs for performing atask The probability of resource acquirement is denoted as119875Resource and the cost is denoted as Costresource This action isrepresented as Res
The candidate action set of an agent is shown in Table 2
42 Task Workflow The purpose of terrorist organization isto produce terror events as much as possible and we simplifythe event to a task in this model The requirement of a taskis human and resource and they are simplified as agentsand resource The general workflow of terror organization ismodeled as the following four steps
Step 1 (generating task) The task is generated by the cellleader of action cell If an action cell is not operating anytask a new task will be created by the cell leader Eachtask 120593 requires amount of resource and agents whichcan be denoted as 120593(resource) and 120593(agent) respectivelyrepresenting the number of resources and agents required tofulfill a task
Step 2 (generating requirement) The requirement is evalu-ated by the cell leader and the element requirements will besent by the agents in the cell Assuming the size of action cell119866119894
AC is |119873119894AC| then the required number of agents is
119873Requirement = 120593 (agent) minus10038161003816100381610038161003816119873119894
AC10038161003816100381610038161003816 (1)
where119873Requirement is the number of generated human require-ments The number of required resources is 120593(resource) Theprocesses of human supplement and resource supplement aredifferent The human requirement will be directly sent to thesuperior and for the resource the leader will interact with allthemembers If there is a member connecting to the resource
cell the requirement will be sent to the member otherwiseit will be sent to the superior
Step 3 (agent response) After the requirements are sent theagent who receives the message will process the requirementbased on its role and relation Figure 3 shows the responsemechanism of an agent to a requirement The agent interactswith the other agent and it can get the information about theother agent In this response process the type of the link builtbetween two members in different cells is Relpp
Step 4 (task execution) Once the requirements are fulfilledthe action cell will launch an event and this is done by thecell leader of action cell The activity circle of action cell canbe illustrated as Figure 4 We hypothesize that the memberswill be lost in performing the task The probability for eachmember to be arrested is denoted as 119875arrest which dependson the efficiency of counterterror department
43 Organizational Recovery Mechanism The recovery pro-cess of terrorist network is the key mechanism for organiza-tion to survive Resilience is a dynamic process associatedwith systems that persist and perform their primary tasksunder pressure from exogenous shocks [32] We do notconsider the case of large-scale attack to the terrorist networkand focus on the normal consumption during the operationalprocess (a member can also be lost if he changes his mindwhich may happen because of the decay of his faith inthe organization In order to focus on the organizationaldynamics we hypothesize that the members will not be lostbecause of the belief issue) The loss in the task concernstwo types of cells the action cell and the resource cell Therecovery process is triggered after a task is performed Ourhypothesis is that the recovery mechanism comprises twobehaviors
(1) Check the leader of the action cell If the leader ofthe action cell is lost a new leader will be generatedfrom the remaining members The selection rule isbased on the attribute Age that is the agent whostays the longest in the cell will be designated as thenew leader This is because when an individual stayslonger in the organization it usually has the richestexperience (expertise knowledge etc) which is themost important characteristic of an action leader
(2) After that the new leader tries to build a new linkto the superior cell with the probability 119875reconnect Thetype of the newly built link is RelssThen the recoveryaction is finished
Figure 5 displays the overall recovery process of anaction cell It can be found that the recovery mechanismof the organization is partly finished by the organizationaloperational process
Resource cell also has recovery mechanism as the mem-ber may be lost in the process of performing a task In themodel we only consider the operational consumption andhypothesize that the member is only lost when the agent in
6 Discrete Dynamics in Nature and Society
If it is a member ofrequired cell
Find a neighborthat connects to the
required cell
Pass the requirement tothe agent
No
Yes
Pass the requirement tothe superior
No
If it can fulfil therequirement
Yes
Resource transferring
If it connects to therequiring cell
Yes
Yes
Find a member in thecell that can fulfilthe requirement
No
No
YesBuild a link to the select
agent in the requiring cell
No
Type of requirement
Resource
Human moving
Human
Receive a requirement
Finish a requirement
ActionChoice
Figure 3 The response mechanism of agent to a requirement
Generaterequirements
Fulfil therequirements Perform a task
Recovery Human loss
No loss
Figure 4 The activity circle of action cell
action cell who links to it is arrestedThe recoverymechanismof resource cell is triggered by the event that a member islost When a member is lost then a human requirement isgenerated by the cell leader
As illustrated in Figure 6 if member 1199096 who is in theaction cell and communicates with the resource sender 1199095 isarrested then 1199095 is lost tooThe leader of resource cell 1199091 then
sends human requirement to the conduct cell for supplementso as tomaintain functional completeness Even though somemembers are lost the resource cell can also acquire resourcebut the efficiency is lower than the complete cell
44 Organizational GrowthMechanism The launched eventslead to organizational growth and the correspondingincrease in size leads to faster production of new events[33] The more frequent the terrorist acts the faster thenetwork grows Events and social influence promote theappeal of organization and recruitment of new militants [12]For this consideration we assume that new action cell willbe generated after certain tasks are finished and a threshold120590Task is used as the trigger of the organizational growth Thesteps of growth can be modeled as follows
Step 1 (leader requirement) A human requirement is gener-ated and assigned to the member in the conduct cell who isin charge of action cell
Discrete Dynamics in Nature and Society 7
Lose members
Lose leader Generate new leader Human supplement
Human supplement New action cell
New action cell
Reconnection
Action cellPerform task
Perform task
Cell leaderx1
x1
x1
x1
x2
x2
x2
x2
x2
x3
x3x
3x3
x3
x3
x3
x3
x4
x5
x5
x5
x5
x5
x5 x
5
x5
x6
x6
x7
x7
x8
x8
x9
x9
Figure 5 The recovery process of action cell
Resourcesupplement
Action cell
Memberlost
Action cell
Resource cell Lose member Human supplement New resource cell
x1
x1x
1
x2
x2x
2
x3
x3x
3
x4
x4
x4
x5
x6
x1
x2
x3
x4
x5
x6
x7
x7
Figure 6 The recovery mechanism of resource cell
Step 2 (leader designation) After the target agent moving tothe new cell it is assigned as the leader of the new actioncell In this process the role of the agent is changed from amember of the training cell to the leader of the new cell
Step 3 (cell formation) Although the newly built action cellhas only one agent it can be performed as a cell whichmeansit can be assigned a task and generate the human and resourcerequirements and send it to the other agents The completeaction cell would form across the organizational task flow
The mechanism of organizational growth is shown inFigure 7 When the tasks performed by action cells reacha threshold 120590Task the network grows and establishes a newaction cell First the leader of the new cell (1199092) is assignedby the operational leader (1199091) in conduct cell Secondlyrequirement of human supplement is sent by the new leaderThirdly the members are transferred through the workflowof human supplement until the new action cell is formedSimultaneously the sizes of resource cell recruitment celland training cell all grow in order to satisfy the growingdemand of actions
5 Organizational Performance Evaluations
In order to analyze the organizational dynamics the indica-tors for evaluating performance are constructed based on thestructural model described in Section 3
As mentioned in [34] we evaluate the performance usingcost and benefit The cost consists of CostReP CostTrainingCostRecruit and CostResource These costs can be recordedduring the simulation and the total cost Costtotal can becalculated as
Costtotal = CostReP + CostTraining + CostRecruit
+ CostResource(2)
Organizational Size The most intuitional evaluation of thedevelopment of terrorist network is the organizational sizethat is the total number of agents in the terrorist organiza-tion and it can be calculated as follows
Size119866 = |119866|
=1003816100381610038161003816119866RC1003816100381610038161003816 +1003816100381610038161003816119866Co1003816100381610038161003816 +1003816100381610038161003816119866TC1003816100381610038161003816 +1003816100381610038161003816119866Re1003816100381610038161003816 + sum119894
10038161003816100381610038161003816119866119894
AC10038161003816100381610038161003816
(3)
Action CellsThe number of action cells measures the growthof terrorist organization from the operational aspectThis canbe measured through the value of 119899ac
Organizational Benefits The goal of terrorist organization isto perform attacks or take similar activities to reach politicalpurpose It is intuitional to use the number of completed tasksas the organizational benefits The task set Φ contains all the
8 Discrete Dynamics in Nature and Society
Conduct cell
Action leader
Conduct cell
Leader of newaction cell
Conduct cell Conduct cell
Conduct cell Assign leader Human supplement Generate new action cell
x1
x1x
1
x1
x2
x2
x2
x3
x3
x4
x4
x5
x6
Figure 7 The growth mechanism of terrorist network
tasks that have been performed The benefit can be obtainedusing
Benefit = |Φ| = 119896 (4)
Cost-Benefits Ratio This is to evaluate the cost-benefit of theactivity of performing tasks As the terrorist organizationalactivity is task driven we here only consider the requirementpassing behavior which measures the efficiency of how theterrorist is organizedThe cost-benefit is measured by the rateof cost and benefit which is denoted as CB so the cost-benefitof requirement passing CBReP is calculated as
CBReP =CostRePBenefit (5)
The other cost-benefits CBTraining CBRecruit andCBResource can also be calculated as (5)
6 Experimental Results and Discussions
In order to quantify the organizational dynamics we per-formed the experiments in a simulation environment anddiscussed the performance and the influence factors basedon the results Potential strategies for counterterrorism arediscussed with the results of sensitivity analysis
61 Experimental Setup The initial organizational structureof terrorist network was generated as shown in Figure 2 aswell as the roles of the individuals We tested the organiza-tional dynamics model in an open environment where thenumber of actions at each simulation step was not limitedIn order to carry out a comparative analysis we also useda limited environment with a parameter to represent thelimited cost 120590Cost which was the number of actions that canbe taken at each step in the network
(1) The initial terrorist network included a conduct cellthree action cells a resource cell a recruit cell and atraining cell The simulation step was 1000
(2) At each step cell leaders checked the situation of thecells and took actions based on the action set andworkflows (eg action cell performs the task if allthe requirements are finished) After that themessagelists of the agents who related to the actions wereupdated
(3) The structure of terrorist network was changed basedon the results of the cell actions (lose members linksetc)
(4) At each step the agents updated their message listsaccording to the interactive mechanism describedabove including message lists of themselves and themessage lists of the interactive objectsThe number ofactions which can be taken by an agent at one step isdenoted as 120590Behavior
(5) At each step the organization checks the number ofcompleted tasks If it reached a threshold 120590Task a newaction cell would be generated by the member in theconduct cell This means that a new action cell isgenerated at the time every 120590Task tasks are performed
(6) In the limited environments if the amount of actionsat each step reached a threshold 120590Cost the actions ofthe rest of agents at current step were stopped untilthe next step
For clarity the values of parameters that are used in thisexperiment and their explanations are listed in Table 3
62 Organizational Dynamics
621 Organizational Performance Figure 8 shows the sim-ulation results of organizational performance First theincrease of costs (Figure 8(a)) and tasks (Figure 8(b)) tendsto accelerate over time and this is because of the growth ofterrorist organization This is consistent with the results ofempirical studies that the violent events tend to acceleratewith increasing size and experience [33] There is a similarlinear relationship between the cost and completed tasksAlthough the cost during the processes of recruiting andresource acquirement depends on the probabilities whichare influenced by randomness the results show statisticallinear relationship It is easy to understand because theorganizational behaviors are driven by task Second thevalues of CostRecruit and CostResource are much higher thanCostTraining and this is caused by the probabilistic behav-iors of recruitment and resource acquirement The sum ofCostTraining CostRecruit and CostResource is much higher thanthat of CostReP and this is consistent with the results ofempirical studies that the preparation of a task is much moreexpensive than performing a task because multiple actionsare required to launch an attack Third Figure 8(a) showsthe differences caused by the task workflows CostRecruit andCostResource are influenced by the probabilities The action
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
6 Discrete Dynamics in Nature and Society
If it is a member ofrequired cell
Find a neighborthat connects to the
required cell
Pass the requirement tothe agent
No
Yes
Pass the requirement tothe superior
No
If it can fulfil therequirement
Yes
Resource transferring
If it connects to therequiring cell
Yes
Yes
Find a member in thecell that can fulfilthe requirement
No
No
YesBuild a link to the select
agent in the requiring cell
No
Type of requirement
Resource
Human moving
Human
Receive a requirement
Finish a requirement
ActionChoice
Figure 3 The response mechanism of agent to a requirement
Generaterequirements
Fulfil therequirements Perform a task
Recovery Human loss
No loss
Figure 4 The activity circle of action cell
action cell who links to it is arrestedThe recoverymechanismof resource cell is triggered by the event that a member islost When a member is lost then a human requirement isgenerated by the cell leader
As illustrated in Figure 6 if member 1199096 who is in theaction cell and communicates with the resource sender 1199095 isarrested then 1199095 is lost tooThe leader of resource cell 1199091 then
sends human requirement to the conduct cell for supplementso as tomaintain functional completeness Even though somemembers are lost the resource cell can also acquire resourcebut the efficiency is lower than the complete cell
44 Organizational GrowthMechanism The launched eventslead to organizational growth and the correspondingincrease in size leads to faster production of new events[33] The more frequent the terrorist acts the faster thenetwork grows Events and social influence promote theappeal of organization and recruitment of new militants [12]For this consideration we assume that new action cell willbe generated after certain tasks are finished and a threshold120590Task is used as the trigger of the organizational growth Thesteps of growth can be modeled as follows
Step 1 (leader requirement) A human requirement is gener-ated and assigned to the member in the conduct cell who isin charge of action cell
Discrete Dynamics in Nature and Society 7
Lose members
Lose leader Generate new leader Human supplement
Human supplement New action cell
New action cell
Reconnection
Action cellPerform task
Perform task
Cell leaderx1
x1
x1
x1
x2
x2
x2
x2
x2
x3
x3x
3x3
x3
x3
x3
x3
x4
x5
x5
x5
x5
x5
x5 x
5
x5
x6
x6
x7
x7
x8
x8
x9
x9
Figure 5 The recovery process of action cell
Resourcesupplement
Action cell
Memberlost
Action cell
Resource cell Lose member Human supplement New resource cell
x1
x1x
1
x2
x2x
2
x3
x3x
3
x4
x4
x4
x5
x6
x1
x2
x3
x4
x5
x6
x7
x7
Figure 6 The recovery mechanism of resource cell
Step 2 (leader designation) After the target agent moving tothe new cell it is assigned as the leader of the new actioncell In this process the role of the agent is changed from amember of the training cell to the leader of the new cell
Step 3 (cell formation) Although the newly built action cellhas only one agent it can be performed as a cell whichmeansit can be assigned a task and generate the human and resourcerequirements and send it to the other agents The completeaction cell would form across the organizational task flow
The mechanism of organizational growth is shown inFigure 7 When the tasks performed by action cells reacha threshold 120590Task the network grows and establishes a newaction cell First the leader of the new cell (1199092) is assignedby the operational leader (1199091) in conduct cell Secondlyrequirement of human supplement is sent by the new leaderThirdly the members are transferred through the workflowof human supplement until the new action cell is formedSimultaneously the sizes of resource cell recruitment celland training cell all grow in order to satisfy the growingdemand of actions
5 Organizational Performance Evaluations
In order to analyze the organizational dynamics the indica-tors for evaluating performance are constructed based on thestructural model described in Section 3
As mentioned in [34] we evaluate the performance usingcost and benefit The cost consists of CostReP CostTrainingCostRecruit and CostResource These costs can be recordedduring the simulation and the total cost Costtotal can becalculated as
Costtotal = CostReP + CostTraining + CostRecruit
+ CostResource(2)
Organizational Size The most intuitional evaluation of thedevelopment of terrorist network is the organizational sizethat is the total number of agents in the terrorist organiza-tion and it can be calculated as follows
Size119866 = |119866|
=1003816100381610038161003816119866RC1003816100381610038161003816 +1003816100381610038161003816119866Co1003816100381610038161003816 +1003816100381610038161003816119866TC1003816100381610038161003816 +1003816100381610038161003816119866Re1003816100381610038161003816 + sum119894
10038161003816100381610038161003816119866119894
AC10038161003816100381610038161003816
(3)
Action CellsThe number of action cells measures the growthof terrorist organization from the operational aspectThis canbe measured through the value of 119899ac
Organizational Benefits The goal of terrorist organization isto perform attacks or take similar activities to reach politicalpurpose It is intuitional to use the number of completed tasksas the organizational benefits The task set Φ contains all the
8 Discrete Dynamics in Nature and Society
Conduct cell
Action leader
Conduct cell
Leader of newaction cell
Conduct cell Conduct cell
Conduct cell Assign leader Human supplement Generate new action cell
x1
x1x
1
x1
x2
x2
x2
x3
x3
x4
x4
x5
x6
Figure 7 The growth mechanism of terrorist network
tasks that have been performed The benefit can be obtainedusing
Benefit = |Φ| = 119896 (4)
Cost-Benefits Ratio This is to evaluate the cost-benefit of theactivity of performing tasks As the terrorist organizationalactivity is task driven we here only consider the requirementpassing behavior which measures the efficiency of how theterrorist is organizedThe cost-benefit is measured by the rateof cost and benefit which is denoted as CB so the cost-benefitof requirement passing CBReP is calculated as
CBReP =CostRePBenefit (5)
The other cost-benefits CBTraining CBRecruit andCBResource can also be calculated as (5)
6 Experimental Results and Discussions
In order to quantify the organizational dynamics we per-formed the experiments in a simulation environment anddiscussed the performance and the influence factors basedon the results Potential strategies for counterterrorism arediscussed with the results of sensitivity analysis
61 Experimental Setup The initial organizational structureof terrorist network was generated as shown in Figure 2 aswell as the roles of the individuals We tested the organiza-tional dynamics model in an open environment where thenumber of actions at each simulation step was not limitedIn order to carry out a comparative analysis we also useda limited environment with a parameter to represent thelimited cost 120590Cost which was the number of actions that canbe taken at each step in the network
(1) The initial terrorist network included a conduct cellthree action cells a resource cell a recruit cell and atraining cell The simulation step was 1000
(2) At each step cell leaders checked the situation of thecells and took actions based on the action set andworkflows (eg action cell performs the task if allthe requirements are finished) After that themessagelists of the agents who related to the actions wereupdated
(3) The structure of terrorist network was changed basedon the results of the cell actions (lose members linksetc)
(4) At each step the agents updated their message listsaccording to the interactive mechanism describedabove including message lists of themselves and themessage lists of the interactive objectsThe number ofactions which can be taken by an agent at one step isdenoted as 120590Behavior
(5) At each step the organization checks the number ofcompleted tasks If it reached a threshold 120590Task a newaction cell would be generated by the member in theconduct cell This means that a new action cell isgenerated at the time every 120590Task tasks are performed
(6) In the limited environments if the amount of actionsat each step reached a threshold 120590Cost the actions ofthe rest of agents at current step were stopped untilthe next step
For clarity the values of parameters that are used in thisexperiment and their explanations are listed in Table 3
62 Organizational Dynamics
621 Organizational Performance Figure 8 shows the sim-ulation results of organizational performance First theincrease of costs (Figure 8(a)) and tasks (Figure 8(b)) tendsto accelerate over time and this is because of the growth ofterrorist organization This is consistent with the results ofempirical studies that the violent events tend to acceleratewith increasing size and experience [33] There is a similarlinear relationship between the cost and completed tasksAlthough the cost during the processes of recruiting andresource acquirement depends on the probabilities whichare influenced by randomness the results show statisticallinear relationship It is easy to understand because theorganizational behaviors are driven by task Second thevalues of CostRecruit and CostResource are much higher thanCostTraining and this is caused by the probabilistic behav-iors of recruitment and resource acquirement The sum ofCostTraining CostRecruit and CostResource is much higher thanthat of CostReP and this is consistent with the results ofempirical studies that the preparation of a task is much moreexpensive than performing a task because multiple actionsare required to launch an attack Third Figure 8(a) showsthe differences caused by the task workflows CostRecruit andCostResource are influenced by the probabilities The action
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Discrete Dynamics in Nature and Society 7
Lose members
Lose leader Generate new leader Human supplement
Human supplement New action cell
New action cell
Reconnection
Action cellPerform task
Perform task
Cell leaderx1
x1
x1
x1
x2
x2
x2
x2
x2
x3
x3x
3x3
x3
x3
x3
x3
x4
x5
x5
x5
x5
x5
x5 x
5
x5
x6
x6
x7
x7
x8
x8
x9
x9
Figure 5 The recovery process of action cell
Resourcesupplement
Action cell
Memberlost
Action cell
Resource cell Lose member Human supplement New resource cell
x1
x1x
1
x2
x2x
2
x3
x3x
3
x4
x4
x4
x5
x6
x1
x2
x3
x4
x5
x6
x7
x7
Figure 6 The recovery mechanism of resource cell
Step 2 (leader designation) After the target agent moving tothe new cell it is assigned as the leader of the new actioncell In this process the role of the agent is changed from amember of the training cell to the leader of the new cell
Step 3 (cell formation) Although the newly built action cellhas only one agent it can be performed as a cell whichmeansit can be assigned a task and generate the human and resourcerequirements and send it to the other agents The completeaction cell would form across the organizational task flow
The mechanism of organizational growth is shown inFigure 7 When the tasks performed by action cells reacha threshold 120590Task the network grows and establishes a newaction cell First the leader of the new cell (1199092) is assignedby the operational leader (1199091) in conduct cell Secondlyrequirement of human supplement is sent by the new leaderThirdly the members are transferred through the workflowof human supplement until the new action cell is formedSimultaneously the sizes of resource cell recruitment celland training cell all grow in order to satisfy the growingdemand of actions
5 Organizational Performance Evaluations
In order to analyze the organizational dynamics the indica-tors for evaluating performance are constructed based on thestructural model described in Section 3
As mentioned in [34] we evaluate the performance usingcost and benefit The cost consists of CostReP CostTrainingCostRecruit and CostResource These costs can be recordedduring the simulation and the total cost Costtotal can becalculated as
Costtotal = CostReP + CostTraining + CostRecruit
+ CostResource(2)
Organizational Size The most intuitional evaluation of thedevelopment of terrorist network is the organizational sizethat is the total number of agents in the terrorist organiza-tion and it can be calculated as follows
Size119866 = |119866|
=1003816100381610038161003816119866RC1003816100381610038161003816 +1003816100381610038161003816119866Co1003816100381610038161003816 +1003816100381610038161003816119866TC1003816100381610038161003816 +1003816100381610038161003816119866Re1003816100381610038161003816 + sum119894
10038161003816100381610038161003816119866119894
AC10038161003816100381610038161003816
(3)
Action CellsThe number of action cells measures the growthof terrorist organization from the operational aspectThis canbe measured through the value of 119899ac
Organizational Benefits The goal of terrorist organization isto perform attacks or take similar activities to reach politicalpurpose It is intuitional to use the number of completed tasksas the organizational benefits The task set Φ contains all the
8 Discrete Dynamics in Nature and Society
Conduct cell
Action leader
Conduct cell
Leader of newaction cell
Conduct cell Conduct cell
Conduct cell Assign leader Human supplement Generate new action cell
x1
x1x
1
x1
x2
x2
x2
x3
x3
x4
x4
x5
x6
Figure 7 The growth mechanism of terrorist network
tasks that have been performed The benefit can be obtainedusing
Benefit = |Φ| = 119896 (4)
Cost-Benefits Ratio This is to evaluate the cost-benefit of theactivity of performing tasks As the terrorist organizationalactivity is task driven we here only consider the requirementpassing behavior which measures the efficiency of how theterrorist is organizedThe cost-benefit is measured by the rateof cost and benefit which is denoted as CB so the cost-benefitof requirement passing CBReP is calculated as
CBReP =CostRePBenefit (5)
The other cost-benefits CBTraining CBRecruit andCBResource can also be calculated as (5)
6 Experimental Results and Discussions
In order to quantify the organizational dynamics we per-formed the experiments in a simulation environment anddiscussed the performance and the influence factors basedon the results Potential strategies for counterterrorism arediscussed with the results of sensitivity analysis
61 Experimental Setup The initial organizational structureof terrorist network was generated as shown in Figure 2 aswell as the roles of the individuals We tested the organiza-tional dynamics model in an open environment where thenumber of actions at each simulation step was not limitedIn order to carry out a comparative analysis we also useda limited environment with a parameter to represent thelimited cost 120590Cost which was the number of actions that canbe taken at each step in the network
(1) The initial terrorist network included a conduct cellthree action cells a resource cell a recruit cell and atraining cell The simulation step was 1000
(2) At each step cell leaders checked the situation of thecells and took actions based on the action set andworkflows (eg action cell performs the task if allthe requirements are finished) After that themessagelists of the agents who related to the actions wereupdated
(3) The structure of terrorist network was changed basedon the results of the cell actions (lose members linksetc)
(4) At each step the agents updated their message listsaccording to the interactive mechanism describedabove including message lists of themselves and themessage lists of the interactive objectsThe number ofactions which can be taken by an agent at one step isdenoted as 120590Behavior
(5) At each step the organization checks the number ofcompleted tasks If it reached a threshold 120590Task a newaction cell would be generated by the member in theconduct cell This means that a new action cell isgenerated at the time every 120590Task tasks are performed
(6) In the limited environments if the amount of actionsat each step reached a threshold 120590Cost the actions ofthe rest of agents at current step were stopped untilthe next step
For clarity the values of parameters that are used in thisexperiment and their explanations are listed in Table 3
62 Organizational Dynamics
621 Organizational Performance Figure 8 shows the sim-ulation results of organizational performance First theincrease of costs (Figure 8(a)) and tasks (Figure 8(b)) tendsto accelerate over time and this is because of the growth ofterrorist organization This is consistent with the results ofempirical studies that the violent events tend to acceleratewith increasing size and experience [33] There is a similarlinear relationship between the cost and completed tasksAlthough the cost during the processes of recruiting andresource acquirement depends on the probabilities whichare influenced by randomness the results show statisticallinear relationship It is easy to understand because theorganizational behaviors are driven by task Second thevalues of CostRecruit and CostResource are much higher thanCostTraining and this is caused by the probabilistic behav-iors of recruitment and resource acquirement The sum ofCostTraining CostRecruit and CostResource is much higher thanthat of CostReP and this is consistent with the results ofempirical studies that the preparation of a task is much moreexpensive than performing a task because multiple actionsare required to launch an attack Third Figure 8(a) showsthe differences caused by the task workflows CostRecruit andCostResource are influenced by the probabilities The action
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
8 Discrete Dynamics in Nature and Society
Conduct cell
Action leader
Conduct cell
Leader of newaction cell
Conduct cell Conduct cell
Conduct cell Assign leader Human supplement Generate new action cell
x1
x1x
1
x1
x2
x2
x2
x3
x3
x4
x4
x5
x6
Figure 7 The growth mechanism of terrorist network
tasks that have been performed The benefit can be obtainedusing
Benefit = |Φ| = 119896 (4)
Cost-Benefits Ratio This is to evaluate the cost-benefit of theactivity of performing tasks As the terrorist organizationalactivity is task driven we here only consider the requirementpassing behavior which measures the efficiency of how theterrorist is organizedThe cost-benefit is measured by the rateof cost and benefit which is denoted as CB so the cost-benefitof requirement passing CBReP is calculated as
CBReP =CostRePBenefit (5)
The other cost-benefits CBTraining CBRecruit andCBResource can also be calculated as (5)
6 Experimental Results and Discussions
In order to quantify the organizational dynamics we per-formed the experiments in a simulation environment anddiscussed the performance and the influence factors basedon the results Potential strategies for counterterrorism arediscussed with the results of sensitivity analysis
61 Experimental Setup The initial organizational structureof terrorist network was generated as shown in Figure 2 aswell as the roles of the individuals We tested the organiza-tional dynamics model in an open environment where thenumber of actions at each simulation step was not limitedIn order to carry out a comparative analysis we also useda limited environment with a parameter to represent thelimited cost 120590Cost which was the number of actions that canbe taken at each step in the network
(1) The initial terrorist network included a conduct cellthree action cells a resource cell a recruit cell and atraining cell The simulation step was 1000
(2) At each step cell leaders checked the situation of thecells and took actions based on the action set andworkflows (eg action cell performs the task if allthe requirements are finished) After that themessagelists of the agents who related to the actions wereupdated
(3) The structure of terrorist network was changed basedon the results of the cell actions (lose members linksetc)
(4) At each step the agents updated their message listsaccording to the interactive mechanism describedabove including message lists of themselves and themessage lists of the interactive objectsThe number ofactions which can be taken by an agent at one step isdenoted as 120590Behavior
(5) At each step the organization checks the number ofcompleted tasks If it reached a threshold 120590Task a newaction cell would be generated by the member in theconduct cell This means that a new action cell isgenerated at the time every 120590Task tasks are performed
(6) In the limited environments if the amount of actionsat each step reached a threshold 120590Cost the actions ofthe rest of agents at current step were stopped untilthe next step
For clarity the values of parameters that are used in thisexperiment and their explanations are listed in Table 3
62 Organizational Dynamics
621 Organizational Performance Figure 8 shows the sim-ulation results of organizational performance First theincrease of costs (Figure 8(a)) and tasks (Figure 8(b)) tendsto accelerate over time and this is because of the growth ofterrorist organization This is consistent with the results ofempirical studies that the violent events tend to acceleratewith increasing size and experience [33] There is a similarlinear relationship between the cost and completed tasksAlthough the cost during the processes of recruiting andresource acquirement depends on the probabilities whichare influenced by randomness the results show statisticallinear relationship It is easy to understand because theorganizational behaviors are driven by task Second thevalues of CostRecruit and CostResource are much higher thanCostTraining and this is caused by the probabilistic behav-iors of recruitment and resource acquirement The sum ofCostTraining CostRecruit and CostResource is much higher thanthat of CostReP and this is consistent with the results ofempirical studies that the preparation of a task is much moreexpensive than performing a task because multiple actionsare required to launch an attack Third Figure 8(a) showsthe differences caused by the task workflows CostRecruit andCostResource are influenced by the probabilities The action
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Discrete Dynamics in Nature and Society 9
Table 3 Parameters setting
Parameters Values Explanations119899ac 3 The initial number of action cells1003816100381610038161003816119873AC1003816100381610038161003816 5 The number of agents in an action cell
1003816100381610038161003816119873Co1003816100381610038161003816 4 The number of agents in a conduct cell
1003816100381610038161003816119873RC1003816100381610038161003816 6 The initial number of agents in a resource
cell1003816100381610038161003816119873TC1003816100381610038161003816 6 The initial number of agents in a training
cell1003816100381610038161003816119873Re1003816100381610038161003816 6 The initial number of agents in a recruit
cellStep 1000 The number of simulation steps119875Recruit 01 The probability of recruitment119875Resource 01 The probability of resource acquirement119875arrest 01 The probability of arrest119875reconnect 01 The probability of reconnection
120590Behavior 3 The number of behaviors that can betaken by each agent in one step
120590Training 20 The time of training a member
120590Task 50 The threshold of tasks completed forbuilding new action cell
120590Cost 30 The threshold of cost limited at each step(in limited environment)
120593(agent) 5 The number of humans required toperform a task
120593(agent) 1 The number of resources required toperform a task
may need to be taken many times before its success andit causes higher cost On the other hand with low arrestprobability the number of recruit requirements is smallerthan the number of resource requirements thus CostRecruitis lower than CostResource Fourth because of the growthmechanism and human consumption the number of agentsshows a trend of fluctuating growth (Figure 8(c)) During thisprocess the mechanisms of network growth and recovery arethe main factors for terrorist network to grow Comparedwith the agent number the cell number is more stable(Figure 8(d))
Figure 8 has shown the general performance of the terror-ist organization in a given environment For further analysisof the organizational dynamics the total cost Costtotal at eachstep is given in Figure 9 Despite overall growth of perfor-mance the number of actions at each step is fluctuating Thecurve shows that the activities of terrorists are very irregulareven in the case of regular working processes Besides theorganizational activity can be very elastic and this indicatesthat the task driven mechanism makes the terrorists do notneed to act all the time which is an important requirementof security for terrorist organizationThefluctuating curves ofactivity at each step and its corresponding organizational sizeshow that the general SNA method may be ineffective whenanalyzing the dynamics of terrorist network
Figure 10 shows the results of organizational cost-benefitratio It is interesting that the four curves are fluctuating
and increasing and finally reach relative stable values Par-ticularly CBReP shows the same trend with the other cost-benefit ratios At the beginning of the simulation becauseof the low requirement and initial state of the organization(resource human etc) all the costs are relatively low Withthe increase of requirement the costs are increasing andfinally reach a relatively stable state This trend impliesthat in an environment of persistence activity the cost ofperforming a task is stable although the structure of theorganization changes (building new links)This is the result ofcooperation by the terrorists according to the organizationalprocess As a system the performance of the organizationis limited by the overall input but the efficiency is deter-mined by the interaction between the individuals and corre-sponding behaviors (building new link efficient requirementpassing)
622 Organizational Performance in a Limited EnvironmentFigures 11(a) and 11(b) show that the cost limitation at eachstep can slow down the speed of task growth The influenceis not obvious at the beginning With the growth of terroristnetwork the number of actions to be taken at each step isincreasing (Figure 11(b)) The limitation reduces the numberof actions which slows down the process of workflows toreduce the number of completed tasks
This result means that the limitation of the organizationalactivities can reduce the terror events by limiting the effi-ciency of organization which leads to the drastic fluctuatingof the agent number as in Figure 11(b) compared withFigure 8(c) The reason of this result is that the assumptionspeed is higher than the recovery speed As the action cellcannot recover timely the period of performing a task isgetting longer which leads to the overall decrease of thecompleted task
Towards the cost-benefit ratio Figure 11(d) shows similartrend as in Figure 10 This means that the limitation of orga-nization activity can only slow down the organization processbut has no effect on the task assumption This is becausethe limitation operation has an effect on the organizationalstructure which means that the external environment ofthe organization determines the organizational productionand the internal structure and mechanism determine theorganizational efficiency
63 Sensitivity Analysis and Potential Strategies Differentvalues of the parameters change the operational environ-ments of terrorist network The organizational dynamics canbe tested in various circumstances and the influence factorscan be analyzed for strategy construction The four probabil-ities parameters are built to model the interaction betweenthe organization and the environment The three thresholdparameters are built to control the internal organizationalprocesses We here focus on the number of completed tasksand the cost-benefit ratio (CBReP) as well as how the changesof parameters affect the performance This part presentsthe operational validation of the simulation experimentsincluding the sensitivity analysis and randomness effectsThepotential strategies are also discussed
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
10 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
2000
4000
6000
8000
10000
12000
The n
umbe
r of c
osts
Training costRecruitment cost
Resource costRequirement cost
(a)
200 400 600 800 10000Simulation step
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s(b)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20
25
The n
umbe
r of a
ctio
n ce
lls in
org
aniz
atio
n
200 400 600 800 10000Simulation step
(d)
Figure 8 Simulation results of organizational performance (a) the cost during the simulation period (b) the number of completed tasks (c)the number of agents in the organization (d) and the number of action cells in the organization
631 Parameters Sensitivity and Potential StrategiesFigure 12 shows the experimental results of performancewithdifferent probabilities of parameters For each configurationthe experiment is repeated 10 times and the results atsimulation step 1000 are recorded
The results in Figures 12(a) 12(b) 12(g) and 12(h)show no apparent difference on the two evaluations whichmeans that the two parameters do not significantly affect theperformance For a growing organization the efficiencies ofrecruitment and reconnection do not significantly affect thenumber of completed tasks This result suggests that witha low probability of arrest (01) the strategy of interveningin organizational recruitment would not significantly reduce
the activities of terrorists For the resource acquirement asshown in Figure 12(c) there is no significant influence unlessthe probability is set to a very low value (01) This meansthat unless the strategy of resource controlling can reducethe probability of resource acquirement to a very low valuethe organizational performance will not be significantlyinfluenced The results in Figures 12(e) and 12(f) indicatethat the difference of arrest probability considerably affectsthe performance When the other conditions are invariableimproving the efficiency of arrest criminals may be the mosteffective way to reduce the performance of terrorist networkAlthough it does not help to prevent the events that havehappened it significantly reduces the growth and violent
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Discrete Dynamics in Nature and Society 11
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
Figure 9 The number of costs at each step
200 400 600 800 1000Simulation step
0
5
10
15
20
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
Figure 10 Simulation results of organizational cost-benefit ratio
events in a long time even more effective than the strategiesof reducing the abilities of terrorists to recruit or to acquireresource Another significant result is that the cost-benefitratio is increasing which means the organization would bemuch harder to perform a task in a high arrest probability
Besides probability parameters the influences of thresh-olds parameters are also analyzed with results shown inFigure 13 It is interesting that the values of the three param-eters all have significant effects on the number of completedtasks A terrorist organization grows faster if it performsmoretasks with high efficiency Another notable observation isthat the cost-benefit ratio of 120590behavior and 120590Training scarcelychanged This result means that even though the completedtasks are changed the ability of the terrorist organizationto perform a task is still unchanged (Figures 13(b) and
Table 4 Statistical significance of different probabilities
Probability 119875Recruit 119875Resource 119875arrest 119875reconnect
01ndash02 0916 0000lowast 0000lowastlowast 0009lowast
02ndash03 0495 0097lowast 0000lowastlowast 0148lowast
03ndash04 0710 0016lowast 0000lowastlowast 0728lowast
04ndash05 0644 0135lowast 0000lowastlowast 0575lowast
05ndash06 0808 0926lowast 0000lowastlowast 0100lowast
06ndash07 0444 0341lowast 0000lowastlowast 0358lowast
07ndash08 0138 0196lowast 0000lowastlowast 0216lowast
08ndash09 0050 0971lowast 0000lowastlowast 0650lowast
09ndash10 0151 0485lowast 0001lowastlowast 0765lowastlowast119875 lt 001 lowast119875 lt 0001
13(d)) In Figures 13(e) and 13(f) not only does the numberof completed tasks decrease but also the cost-benefit ratiodecreases This indicates that the task threshold not onlycan slow down the growth of the organization but also canimprove the efficiency of group
For intervention strategies it is easy to find that thestrategies targeting at these parameters are all effectiveStrengthening the security environment can reduce the activefrequency of terrorists which means that the behaviors ofagent in a step (120590Behavior) will be reduced Intervention onthe interaction of terrorist will prolong the time of trainingan agent (120590Training) Reducing the social influence of terroristattacks would raise the threshold of building new actioncell (120590Task) From the perspective of organizational dynamicsthese strategies may be more effective than finding potentialterrorists to reduce the recruitment of terrorist organization(as shown in Figure 12(a))
For further investigation of the task performance theexperiments with different arrest probability are performedand the results are shown in Figure 14 With the increaseof arrest probability CBTra CBrec and CBReP are increasingThis is because the probability is related to the organiza-tional structure and task performance Another result isthat CostReP becomes higher than CostRec this means thatwhen the arrest probability reaches a level (in this case 03)the highest assumption of the organization would be therequirement passing which means that the main cost forperforming task is the individual interaction
632 Statistical Tests of Randomness Effects In order toexamine the extent to which the randomness affects thesimulation results the experiment was tested following aprocedure of factorial analysis This section presents thestatistical significance of the experiments performed inSection 631 and the results are shown in Tables 4 and 5(119875 lt 005 lowast for 119875 lt 001 and lowastlowast for 119875 lt 0001) In thisanalysis the value used for test is the number of completedtasks in the simulation process The consistency between thissection and previous results proved that the difference of theexperiments is caused by themodel not the randomness andthe previous analysis is meaningful
Table 4 shows the results of statistical significance withdifferent probabilities The test is carried out between two
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
12 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
The n
umbe
r of c
ompl
eted
task
s
200 400 600 800 10000Simulation step
(a)
20
40
60
80
100
120
140
160
180
The n
umbe
r of a
gent
s in
the o
rgan
izat
ion
200 400 600 800 10000Simulation step
(b)
0
10
20
30
40
50
60
70
The n
umbe
r of c
osts
at ea
ch st
ep
200 400 600 800 10000Simulation step
(c)
0
5
10
15
20Th
e cos
t-ben
efit r
atio
200 400 600 800 10000Simulation step
TrainingRecruitment
ResourceRequirement passing
(d)
Figure 11 The organizational performance in a limited environment
Table 5 Statistical significance of different thresholds
Number 120590Behavior 120590Training 120590Task
1 0000lowastlowast (1 2) 0063lowast (10 20) 0000lowastlowast (30 40)2 0000lowastlowast (2 3) 0091lowast (20 30) 0000lowastlowast (40 50)3 0000lowastlowast (3 4) 0000lowastlowast (30 40) 0000lowastlowast (50 60)4 0000lowastlowast (4 5) 0000lowastlowast (40 50) 0000lowastlowast (60 70)lowast119875 lt 001 lowast119875 lt 0001
adjacent probabilities For example in the probability row01ndash02 the value in 119875Recruit column means that it is the
statistical significance result between the group 119875Recruit =01 and the group 119875Recruit = 02 The values in the tableshow the same conclusions discussed in Figure 12The changeof arrest probability significantly affects the organizationalperformance For the probabilities of resource acquirementand reconnection only in the case of low probabilities (01ndash02) the changes of probabilities can significantly affect theresults
Table 5 shows the test results with different values ofthresholds The results are calculated between every twoadjacent values of the thresholds Since in Section 631 eachparameter is tested with 5 values the statistical significance
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Discrete Dynamics in Nature and Society 13
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(a)
0 02 03 04 05 06 07 08 09 1001The probability of recruit
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(b)
0 02 03 04 05 06 07 08 09 1001The probability of resource acquirement
0
200
400
600
800
1000
1200
1400
1600
1800
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of resource acquirement
(d)
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of arrest
(e)
0
5
10
15
20
25
30
35
40
The n
umbe
r of c
ost-b
enefi
t rat
ios
02 03 04 05 06 07 08 09 1001The probability of arrest
(f)
Figure 12 Continued
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
14 Discrete Dynamics in Nature and Society
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
02 03 04 05 06 07 08 09 1001The probability of reconnection
(g)
02 03 04 05 06 07 08 09 1001The probability of reconnection
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
(h)
Figure 12 Simulation results of organizational performance with different (a)-(b) recruit probability (c)-(d) resource probability (e)-(f)arrest probability and (g)-(h) reconnection probability
is carried out between two adjacent values The values inparentheses are the two values of the threshold parameterFrom the results it is easy to find that they show consistentconclusions in Figure 13 Except when 120590Training is low (10ndash20 20ndash30) the results of the other groups are all statisticallysignificant
7 Conclusion
This paper studied the organizational dynamics of terroristnetwork by incorporating a hierarchical cellular networkmodel and mechanisms of organizational processes Mod-eling and simulating the dynamic adaptive behaviors ofterrorists enable us to quantitatively investigate the emergentdynamics of organization and to find the influencing factorsfor constructing intervention strategies Our main conclu-sions are summarized as follows
(1) For a continuous active terrorist organization theorganizational performance is influenced by boththe interactional environment and the organizationalmechanism The efficiency of the organization willbe fluctuating and finally reach a relative stable stateand the state is determined by both the external inputand internal structure andmechanismThis shows thecharacteristics of terrorist organization as a complexsystem
(2) Even in a regular organizational process the activityof the terrorist is irregular (as Figure 9 shows) Thesize of the organization shows a trend of fluctuatinggrowth under the organizational mechanisms andconsumption The recovery mechanism enables theorganization to survive and grow in case of losingmembers
(3) In a limited operational environment the organiza-tion shows more uncertainty in the organization sizeand growth (as in Figure 11(b))This environment canonly ldquoslow downrdquo the process of organizational devel-opment and the assumption for task is unchangedand this is because the limited environment cannotchange the organizational structure and process
(4) For counterterrorism the most effective strategyshould be those which have effect on both the orga-nizational structure and the organizational processFrom this perspective the strategywhich can increasethe arrest rate or reduce the frequency of buildingnew action cell will be the most useful The otherstrategy which has effect on the active frequencyof the terrorist the training time and the resourceacquirement can only reduce the launched events notthe efficiency of the organization
In conclusion this work gives a new insight into organiza-tional dynamics of terrorist network by using ABM methodIt can be used to test various simulation experiments toinvestigate organizational dynamics under different scenar-ios Based on the simulation framework different strategiescan be examined for making counterterrorism policy Afuture development of this research would be to apply thismethod to study the evolution of terrorist network underdifferent intervention strategies
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Discrete Dynamics in Nature and Society 15
0
500
1000
1500
2000
2500
3000
3500
4000
4500Th
e num
ber o
f com
plet
ed ta
sks
2 3 4 51The number of behaviors in one step
(a)
0
1
2
3
4
5
6
7
8
9
10
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The number of behaviors in one step
(b)
20 30 40 5010The time of training an agent
0
200
400
600
800
1000
1200
The n
umbe
r of c
ompl
eted
task
s
(c)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
2 3 4 51The time of training an agent
(d)
0
200
400
600
800
1000
1200
1400
1600
The n
umbe
r of c
ompl
eted
task
s
40 50 60 7030The threshold of tasks
(e)
0
1
2
3
4
5
6
7
8
9
The n
umbe
r of c
ost-b
enefi
t rat
ios
40 50 60 7030The threshold of tasks
(f)
Figure 13 Simulation results of organizational performance with different (a)-(b) threshold of activities (c)-(d) time of training and (e)-(f)thresholds of tasks
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
16 Discrete Dynamics in Nature and Society
200 400 600 800 10000Simulation step
0
10
20
30
40
50Th
e cos
t-ben
efit r
atio
TrainingRecruitment
ResourceRequirement passing
(a)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(b)
200 400 600 800 10000Simulation step
0
10
20
30
40
50
The c
ost-b
enefi
t rat
io
TrainingRecruitment
ResourceRequirement passing
(c)
Figure 14 Simulation results of cost-benefit ratio under different arrest probability (a) 03 (b) 05 and (c) 07
Acknowledgments
This work is supported by the National Natural ScienceFoundation of China (Grant no 71473263) and the Special-ized Research Fund for the Doctoral Program of HigherEducation of China (Grant no 20134307110020)
References
[1] D Conway Modeling Network Evolution Using Graph MotifsComputing Research Repository (CoRR) 2011
[2] D Garlaschelli A Capocci and G Caldarelli ldquoSelf-organizednetwork evolution coupled to extremal dynamicsrdquo NaturePhysics vol 3 no 11 pp 813ndash817 2007
[3] T A B Snijders C E G Steglich and M SchweinbergerldquoModeling the co-evolution of networks and behaviorrdquo inLongitudinal Models in the Behavioral and Related Sciences KV Montfort J Oud and A Satorra Eds pp 41ndash71 LawrenceErlbaum Associates 2007
[4] F Bergenti E Franchi and A Poggi ldquoAgent-based interpreta-tions of classic networkmodelsrdquoComputational andMathemat-ical Organization Theory vol 19 no 2 pp 105ndash127 2013
[5] M E J Newman ldquoThe structure and function of complexnetworksrdquo SIAM Review vol 45 no 2 pp 167ndash256 2003
[6] PV Fellman andRWright ldquoModeling terrorist networks com-plex systems at themid-rangerdquo in Proceedings of the ComplexityEthics and Creativity Conference London UK 2003
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Discrete Dynamics in Nature and Society 17
[7] R Lindelauf P Borm andHHamers ldquoThe influence of secrecyon the communication structure of covert networksrdquo SocialNetworks vol 31 no 2 pp 126ndash137 2009
[8] M Sageman Understanding Terror Networks University ofPennsylvania Press Philadelphia Pa USA 2004
[9] S P Borgatti ldquoIdentifying sets of key players in a socialnetworkrdquoComputational ampMathematical OrganizationTheoryvol 12 no 1 pp 21ndash34 2006
[10] V E Krebs ldquoMapping networks of terrorist cellsrdquo Connectionsvol 24 no 3 pp 43ndash52 2001
[11] H Chen Intelligence and Security Informatics for InternationalSecurity Springer New York NY USA 2005
[12] P J Phillips ldquoThe life cycle of terrorist organizationsrdquo Interna-tional Advances in Economic Research vol 17 no 4 pp 369ndash3852011
[13] K M Carley ldquoDestabilization of covert networksrdquo Computa-tional and Mathematical Organization Theory vol 12 no 1 pp51ndash66 2006
[14] K M Carley ldquoDynamic network analysisrdquo in Dynamic SocialNetwork Modeling and Analysis Workshop Summary andPapers pp 133ndash145 National Academies Press WashingtonDC USA 2003
[15] A Ilachinski ldquoModelling insurgent and terrorist networks asself-organised complex adaptive systemsrdquo International Journalof Parallel Emergent and Distributed Systems vol 27 no 1 pp45ndash77 2012
[16] L Yilmaz ldquoValidation and verification of social processeswithin agent-based computational organization modelsrdquo Com-putational andMathematical OrganizationTheory vol 12 no 4pp 283ndash312 2006
[17] S Bandini S Manzoni and G Vizzari ldquoAgent based modelingand simulation an informatics perspectiverdquo Journal of ArtificialSocieties and Social Simulation vol 12 no 4 article 4 2009
[18] T A Leweling and M E Nissen ldquoDefining and explor-ing the terrorism field toward an intertheoretic agent-basedapproachrdquo Technological Forecasting and Social Change vol 74no 2 pp 165ndash192 2007
[19] G A Backus and R J GlassAn Agent-BasedModel Componentto a Framework for the Analysis of Terrorist-Group DynamicsSandia National Laboratories Livermore Calif USA 2006
[20] P O Neil ldquoDynamic covert network simulationrdquo in SocialComputing Behavioral-Cultural Modeling and Prediction pp239ndash247 Springer Verlag Berlin Germany 2012
[21] M Tsvetovat and M Latek ldquoDynamics of agent organizationsapplication to modeling irregular warfarerdquo in Multi-Agent-Based Simulation IX vol 5269 of Lecture Notes in ComputerScience pp 60ndash70 Springer Berlin Germany 2009
[22] I-C Moon and K M Carley ldquoModeling and simulatingterrorist networks in social and geospatial dimensionsrdquo IEEEIntelligent Systems vol 22 no 5 pp 40ndash49 2007
[23] A Vespignani ldquoModelling dynamical processes in complexsocio-technical systemsrdquoNature Physics vol 8 no 1 pp 32ndash392012
[24] P A C Duijn V Kashirin and PM A Sloot ldquoThe relative inef-fectiveness of criminal network disruptionrdquo Scientific Reportsvol 4 article 4238 2014
[25] S Helfstein and DWright ldquoCovert or convenient Evolution ofterror attack networksrdquo Journal of Conflict Resolution vol 55no 5 pp 785ndash813 2011
[26] D Jones Understanding the Form Function and Logic ofClandestine Insurgent and Terrorist Networks The First Step in
Effective Counternetwork Operations Joint Special OperationsUniversity 2012
[27] M Tsvetovat and K M Carley Structural Knowledge andSuccess of Anti-Terrorist Activity The Downside of StructuralEquivalence Center for Computational Analysis of Social andOrganizational Systems (CASOS) 2005
[28] G AHenkeHowTerrorist Groups Survive ADynamic NetworkAnalysis Approach to the Resilience of Terrorist OrganizationsSchool of Advanced Military Studies 2009
[29] T L Frantz and K M CarleyA Formal Characterization of Cel-lular Networks CMU-ISRI-05-109 Center for ComputationalAnalysis of Social and Organizational Systems (CASOS) 2005
[30] F Ozgul ldquoClassification of terrorist networks and their keyplayersrdquo in Multidisciplinary Social Networks Research L S-L Wang and J J June Eds vol 473 of Communications inComputer and Information Science pp 145ndash157 Springer BerlinGermany 2014
[31] C Leuprecht and K Hall ldquoWhy terror networks are dissimilarhow structure relates to functionrdquo in Networks and NetworkAnalysis for Defence and Security A J Masys Ed LectureNotes in Social Networks pp 83ndash120 Springer InternationalPublishing Gewerbestrasse Switzerland 2014
[32] R M Bakker J Raab and H B Milward ldquoA preliminarytheory of dark network resiliencerdquo Journal of Policy Analysis andManagement vol 31 no 1 pp 33ndash62 2012
[33] A Clauset and K S Gleditsch ldquoThe developmental dynamicsof terrorist organizationsrdquo PLoS ONE vol 7 no 11 Article IDe48633 2012
[34] D Ye M Zhang and D Sutanto ldquoSelf-organization in an agentnetwork a mechanism and a potential applicationrdquo DecisionSupport Systems vol 53 no 3 pp 406ndash417 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Top Related