Giorgini P., EuroPKI 20041 Filling the gap between Requirements Engineering and Public Key/Trust...

Giorgini P., EuroPKI 2004 1

Filling the gap between Filling the gap between Requirements Engineering andRequirements Engineering andPublic Key/Trust Management Public Key/Trust Management

InfrastructuresInfrastructures

Paolo Giorgini

Department of Information and Comm. Tech.

University of Trento (Italy)

Joint work with

Fabio Massacci, John Mylopoulos, and Nicola Zannone


SummarySummary

• Motivation

• Our approach – Secure aware-Tropos– Case study– Formalization– Axioms– Proprerties– Trust Management Implementation

• Conclusion and future work


Trust Management and PKIsTrust Management and PKIs

• Trust Management and PKIs are hot topics in security research: – sophisticated policy languages, algorithms, and system for

managing security credentials

• Solutions based on public-key cryptography and credential have been shown to be well suited in satisfying the security requirements of distributed systems

• However, there is big gap between solutions and the requirements of the entire system


Security and RequirementsSecurity and Requirements

• No methodologies for linking security policy to the mainstream requirements analysis process

• The usual approach towards the inclusion of security within a system is to identify security requirements after system design

• Security mechanisms have to be fitted into a pre-existing design– may not be able to accommodate them– security requirements can generate conflicts functional

requirements of the system


Our goalOur goal

• There are proposals improving on secure engineering or architectures for trust management, but nobody has proposed a methodology that considers together both these approaches

• We want to introduce a trust management system into the requirements engineering framework

– avoid designing an entire system and then retrofitting a PKI on its top, when it is already to late to make it fits snugly


Our proposalOur proposal

• A process that integrates trust, security and system engineering, using the same concepts and notations used for requirements specification

– Three steps approach:1. Functional Requirements modeling

2. Trust Requirements modeling

3. PKI/trust management implementation

• We use Tropos, an agent-oriented methodology, for requirements modeling and analysis


Tropos MethodologyTropos Methodology• Tropos is an agent-oriented software development

methodology, tailored to describe both the organization and the system itself

• Tropos uses concepts of – Actor

Intentional entity: role, position, agent (human or software)

– Goal (softgoal) Strategic interest of an actor

– TaskParticular course of action that can be executed in order to satisfy a goal

– ResourcePhysical or informational entity (without intentionality)

– Social dependency (between two actors)One actor depends on another to accomplish a goal, execute a task, or

deliver a resource


Security-Aware Tropos Security-Aware Tropos

• Tropos has not been designed with security in mind

• We introduce four new relationships: – Trust ,among two agents and a service– Delegation, among two agents and a service– Ownership, between an agent and a service– Offer, between an agent and a service

• And we refine the methodology by– Define functional dependencies of services among actors– Design a trust model among actors – Identify who owns services and who is able to fulfill them


An illustrative Case StudyAn illustrative Case Study

• A health care IS, in which

– Patient, that depends on the hospital for receiving appropriate health care. Further, patients will refuse to share their data if they do not trust the system or do not have sufficient control over the use of their data;

– Hospital, that provides medical treatment and depends on the patients for having their personal information.

– Clinician, physician of the hospital that provides medical health advice and, whenever needed, provide accurate medical treatment;

– Health Care Authority (HCA) that control and guarantee the fair resources allocation and a good quality of the delivered services.

– Medical Information System (MIS), that, according the current privacy legislation, can share the patients medical data if and only if consent is obtained.


The Functional Requirements ModelThe Functional Requirements Model

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

D: DependencyA: AimS: Service


The Trust Requirements ModelThe Trust Requirements Model



O: OwnershipT: Trust


The Trust Management ImplementationThe Trust Management Implementation



2 forms of Delegation: P: Permission (deleg. for use) G: delegation for Grant


Formalization (1)Formalization (1)

Predicates for the functional requirements model• offers(a,s)

• aims(a,s)

• has(a,s)

• depends(a,b,s1,s2)

Predicates for the trust requirements model• owns(a,s)

• trust(a,b,s1,s2,n) n: trust depth

Predicates for the trust management implementation• fulfills(a,s)

• delGrant(idC,a,b,s1,s2,n) idC: certificate identify n: delegation depth

• permission(idC,a,b,s1,s2)


Formalization (2)Formalization (2)

€

delGChain(A,B,S1,S2) ≡∃k s.t. ∃a1...ak ∃n1...nk−1 ∀ i ∈ 1...k −1[ ]

delGrant(idi,ai,ai+1,S1,S2,ni)∧(a1 = A)∧(ak = B)

⎧ ⎨ ⎩

€

permissionChain(A,C,S1,S2) ≡

(permission(idC,A,C,S1,S2))∨

(∃B delGChain(A,B,S1,S2))∧

permission(idC,B,C,S1,S2))

⎧

⎨ ⎪

⎩ ⎪

A way to see depth is the number of re-delegation; depth 1 means that no re-delegation is allowed, depth N that N-1 further step are allowed


AxiomsAxiomsusing Datalog




Properties Properties

We use the DLV system for automatic verification of security requirements




Negative Authorization (1)Negative Authorization (1)

• We use a closed world policy: the lack of an authorization is interpreted as a negative authorization

• This approach has a major problem in the lack of a given authorization for a given actor does not prevent this user from receiving this authorization later on

• We propose an explicit negative authorization, namely an explicit denial for an actor to access a service

• Negative authorizations are stronger than positive authorizations

• Two predicates:– delDenial(idC,a,b,s,n)– prohibition(idC,a,b,s)and analougsly for positive authorization– delDChain(A,B,S)– prohibitionChain(A,C,S)


Negative Authorization (2)Negative Authorization (2)

Axioms

Properties






Trust Management ImplementationTrust Management Implementation• We use the RT framework (by Li et al.), which provides

policy language, semantics, deduction engine, and pragmatic features

• RT includes a declarative, logic-based semantic foundation based on Datalog, support for vocabulary agreement, strongly-typed credential and policies, and flexible delegation structures

• In RT, an entity is a uniquely identified individual or process

• An entity can issue credentials and make requests

• RT uses the notion of role to represent attributes– Entity.Role


Roles in the RT frameworkRoles in the RT framework

• Only the entity A has the authority to A.R, and A does so by issuing role-definition credentials

• An entity A can define A.R to contain A.R1, another role defined by A– A.RA.R1, means that A defines that R1 dominates R

• A credential A.RB.R is a delegation from A to B of authority over R. This can be used to decentralize the user-role assignment.

• A credential of the form A.RB.R1 can be used to define role-mapping across multiple organizations

• The credential A.RA.R1.R2 states that: A.R contains any B.R2 if A.R1 contains B.


Moving to the RT frameworkMoving to the RT framework

permission(ID,A,B,S1,S2)– A.S1B.S2

delGrant(ID,A,B,S1,S2,N)– A.S1B.r.S2

where B allows to use the service S1 to actors in the role B.r


Example 1Example 1

A patient allows his clinician to read his personal/medical data to provide accurate medical treatment.

permission(id,Pat,Cli,Rec,MedTre):- isClinicianOf(Pat,Cli)^owns(Pat,Rec)

In RT:

Pat.recordAc(read,?F:Pat.record) Pat.clinician.provide(?E:medTre)

Given Pat.recordRec and Pat.clinicianCli, one can conclude that

Pat.recordAc(read,Rec)Cli.provide(?E:medTre)


Example 2Example 2

The Medical Information System allows the clinician to write on his patient records to upgrade them.

permission(id,MIS,Cli,Rec,upgrade(Rec)):- isClinicianOf(Pat,Cli)^owns(Pat,Rec)

In RT

MIS.recordAc(write,?F:Pat.record) Pat.clinician.upgrade(?F:Pat.record)

Given Pat.recordRec and Pat.clinicianCli, one can conclude that

MIS.recordAc(write,Rec)Cli.upgrade(Rec)


Conclusion and future workConclusion and future work

• We have introduced a process that integrates security and requirements engineering– A clear separation of trust and delegation relationship

• Our framework supports the automatic verification of security requirements

• We have defined the trust management implementation of our framework into the RT framework

• Future work – incorporating explicitly roles adding time features– integration with the Formal Tropos tool

Giorgini P., EuroPKI 20041 Filling the gap between Requirements Engineering and Public Key/Trust...

Documents

Transcript of Giorgini P., EuroPKI 20041 Filling the gap between Requirements Engineering and Public Key/Trust...