you_never_surf_alone

19
YOU NEVER SURF ALONE. UBIQUITOUS TRACKING OF USERS’ BROWSING HABITS. Silvia Puglisi, David Rebollo-Monedero and Jordi Forné Department of Telematics Engineering Universitat Politecnica de Catalunya

Transcript of you_never_surf_alone

Page 1: you_never_surf_alone

YOU NEVER SURF ALONE. UBIQUITOUS TRACKING OFUSERS’ BROWSING HABITS.

Silvia Puglisi, David Rebollo-Monedero and Jordi Forné

Department of Telematics Engineering Universitat Politecnica de Catalunya

Page 2: you_never_surf_alone

AGENDABackgroundContributionModelling the user's footprintA metric of similarityExperimental resultsConclusions

Page 3: you_never_surf_alone

BACKGROUNDWhen users surf the web a network of personalisationservices tracks their preferences by following their browsinghabits.

These services combine information from different sources:

profiles and accounts that users create,browser cookies,device fingerprinting and so on.

Page 4: you_never_surf_alone

Online profiling often means collecting enough features todistinguish an individual among a group of candidates.

Knowing certain facts about an individual allows the addnetwork (or an attacker) to verify assumptions and makepredictions.

Page 5: you_never_surf_alone

CONTRIBUTIONS1. Introducing a model of the user online footprint.2. Introducing a measure of similarity between the user

profile and the observed advertising profile.3. Measuring how quickly a user is uniquely identified and

tracked by an advertising network.

Page 6: you_never_surf_alone

MODELLING THE USER’S FOOTPRINTWe model the user’s activity as series of events belonging toa certain identity.

Each event is a document containing different information.

Each document can be seen as a hypermedia document i.e.an object possibly containing graphics, audio, video, plaintext and hyperlinks.

We call the hyperlinks selectors and we use these to buildthe connections between documents.

Page 7: you_never_surf_alone
Page 8: you_never_surf_alone

We aggregate keywords each time the user creates a newevent by visiting a different url.

These keywords constitute the user profile of interests.

Page 9: you_never_surf_alone

With this model we assume that a particular category isweighted according to the number of times this has beencounted in the user profile.

We define the profile of a user um as the Probability MassFunction:

A histogram of relative frequencies of tags across the set oftag categories T.

Page 10: you_never_surf_alone

The profile of an ad is defined as the PMF:

Where q_{n,l} is the percentage of tags belonging to thecategory l which have been assigned to this specificadvertising item.

Page 11: you_never_surf_alone

A METRIC OF SIMILARITYWe consider the third party advertising network to operatelike a recommendation system suggesting products orservices based on the users' preferences.

We assume that the ad server suggest advertising based ona measure of similarity.

Page 12: you_never_surf_alone

We use the 1 − norm between the user and the advertistingprofile as a measurement of how the advertising network istracking the user profile:

Page 13: you_never_surf_alone

EXPERIMENTAL METHODOLOGY

Page 14: you_never_surf_alone

WHY WE CRAWLED TWITTER FEEDS

“https://blog.twitter.com/2014/introducing-the-website-tag-for-remarketing”

Page 15: you_never_surf_alone

EXPERIMENTAL RESULTS

Page 16: you_never_surf_alone

CONCLUSIONSWeb tracking happens very quickly in a few subsequentvisits to websites in a large advertising network.

Hypermedia models are able to capture both connectionsbetween entities as weel as individual events features andmetrics.

Page 17: you_never_surf_alone

FUTURE DEVELOPMENTWe would like to introduce a full set of metrics to measurethe distance between the advertising and the user's profile:

2-norm,KL-Divergence between the advertising profile and theobserved user profile,Fisher information.

Page 18: you_never_surf_alone

We are interested in measuring how social networks sharingbuttons and/or commenting services, included on websites,are able to track users even when these have not signed inwith their account.

We are interested to measure how Privacy EnhancingTechniques affect advertising networks.

Page 19: you_never_surf_alone

"I do not want to live in a world where everything I do and say is recorded. That is not something I amwilling to support or live under."

Edward Snowden