The Economics of Mass The Economics of Mass Surveillance and The Surveillance and The Questionable Value of Questionable Value of Anonymous Anonymous CommunicationsCommunicationsGeorge Danezis and Bettina George Danezis and Bettina WittnebenWittnebenPresented By: Warren PowersPresented By: Warren PowersOverviewOverviewIntroductionIntroductionModelModelObtaining network informationObtaining network informationSurveillanceSurveillanceDiscussion of effectivenessDiscussion of effectivenessConclusionConclusionIntroductionIntroductionParticipants belong to clubsParticipants belong to clubsIf one participant is under surveillance, all If one participant is under surveillance, all information shared and membership is information shared and membership is revealedrevealedQuestionsQuestionsHow many need to be under surveillance?How many need to be under surveillance?Who do we put under surveillance?Who do we put under surveillance?How does anonymity affect target selection?How does anonymity affect target selection?ModelModelPeople and spaces (or clubs)People and spaces (or clubs)RelationshipsRelationshipsWhen people belong to spacesWhen people belong to spacesNo links between peopleNo links between peopleRelationships have strengthRelationships have strengthSymbolizes degree of association between Symbolizes degree of association between person and spaceperson and spaceGraph from set of people to set of Graph from set of people to set of spaces.spaces.Extracting the Network from Extracting the Network from DataDataUsed data from mailing list archivesUsed data from mailing list archivesMapped email address to spaceMapped email address to spaceMapped email to personMapped email to personRelations were created from messages Relations were created from messages to liststo listsEffectiveness of Partial Effectiveness of Partial SurveillanceSurveillanceWhat is revealed?What is revealed?Observing one member of a spaceObserving one member of a spaceAll relationships associated with that spaceAll relationships associated with that spaceChoice of targetChoice of targetThose with highest degreeThose with highest degreeAmong spaces not under surveillanceAmong spaces not under surveillanceRepeated as budget allowsRepeated as budget allowsData Uncovered – Full InfoData Uncovered – Full InfoPartial InformationPartial InformationWe can monitor volume of messagesWe can monitor volume of messagesNot degree or correspondentsNot degree or correspondentsTarget selection more difficultTarget selection more difficultLower return on investmentLower return on investmentData Uncovered – Partial InfoData Uncovered – Partial InfoDiscussionDiscussionThe first model could represent no The first model could represent no anonymityanonymityCan obtain much information with little Can obtain much information with little surveillancesurveillanceThe second model represents some The second model represents some anonymityanonymityNo cover trafficNo cover trafficAnonymized communication is helpful but Anonymized communication is helpful but not perfectnot perfectDiminishing ReturnsDiminishing ReturnsInitial investment provides great returnInitial investment provides great returnAs budget is increased, marginal returns As budget is increased, marginal returns decrease.decrease.Cost per unit of intelligence rapidly increasesCost per unit of intelligence rapidly increasesUseful information may be very costlyUseful information may be very costlyPrivacy violation is highPrivacy violation is highInterception FiguresInterception FiguresWarrants issued vs. number under Warrants issued vs. number under surveillancesurveillanceUK populationUK populationFull information graph usedFull information graph usedFormula for those under surveillanceFormula for those under surveillance (0.5/0.01)X(0.5/0.01)X(0.5/0.01)1849 = 92000 people(0.5/0.01)1849 = 92000 peopleInfo on 50 people revealed for each one Info on 50 people revealed for each one monitoredmonitoredFailure of Adaptive Target Failure of Adaptive Target SelectionSelectionAdaptive strategies are inferior to volume Adaptive strategies are inferior to volume selectionselectionAdaptive StrategiesAdaptive StrategiesHigh known degreeHigh known degreeLikely to have links to undiscovered spacesLikely to have links to undiscovered spacesStructural equivalenceStructural equivalenceHigh know degree and few nodes sharing its High know degree and few nodes sharing its positionpositionStrategy ComparisonStrategy ComparisonTarget Selection for DisruptionTarget Selection for DisruptionWhat if goal is to disrupt network?What if goal is to disrupt network?Remove nodes with highest degreeRemove nodes with highest degreeRemove nodes with high volumeRemove nodes with high volumeSelection results not very differentSelection results not very differentNeed to remove twice as many using Need to remove twice as many using volume info.volume info.ConclusionConclusionInformation is leaked through third partiesInformation is leaked through third partiesA small carefully selected set of nodes A small carefully selected set of nodes reveal a large amount of informationreveal a large amount of informationUnlinkablility is not sufficient, Unlinkablility is not sufficient, unobservability is necessaryunobservability is necessarySurveillance will violate privacy of Surveillance will violate privacy of innocent partiesinnocent partiesFinding guilty parties will be costlyFinding guilty parties will be
View Full Document