Research On Security Of Trust Models Utilizing Game Theory In Pervasive Computing Environments

Interactions between mobile entities in pervasive computing have properties suchas open, peer to peer and dynamic etc and therefore it is needed to adopt trust models togenerate and manage various relations between entities. However, because entities arefrom different organizations and are resource limited, they will not take the predefinedactions of trust models according to their own benefits. Hence the security problemsof trust models occur. For such"passive attack", we study the solution based on gametheory and mechanism design to solve the security problems of trust models in thisthesis. Based on summarizing exist research results in home and abroad, we studyfrom several perspectives and do simulations in details. The obtained results in thisthesis enlarge the scope and contend of the disciple called Economics of InformationSecurity. The concrete research content includes:(1) For the problem of low recommendation acquisition rate in the recommenda-tion acquisition process due to the dropping of middle entities and duck of recommen-dation rendering entity, we propose a utility based incentive mechanism to encouragethese two kinds of entities to guarantee the full success of recommendation acquisition.First, we study the selfish model and the cooperation condition of selfish entities utiliz-ing a non cooperative game called Recommendation Acquisition Game. Then considerthe situation where asking prices are proposed by greedy entities, based on a coopera-tive game called Competition and Section Game, we study how the requestor lowers thetotal paid cost utilizing the competition between several routings. Finally, the simula-tion results show that the proposed incentive mechanism is effective and increase about15% to 30% success rate and lowers the asking prices of cooperators.(2) For the cheating in the recommendation, we proposed a strategy-proof trustmechanism based on VCG mechanism from the perspective of non cooperative game. This mechanism can be adopted by continuous and binary trust value models to stim-ulate truthful reportings. By considering the trust decision as a social section process,we prove that the proposed trust mechanism is a VCG mechanism so that the entitycan only maximize its utility when truthfully reports. Furthermore, utilizing WMC al-gorithm, we propose a weighted VCG mechanism to predict trust more accurately andmoreover, we study the characteristic of payment.(3) For discovering more strategy-proof trust mechanisms, based on the proposedgeneral trust mechanism, we study the characteristic properties of incentive compatibleso that all the trust models satisfying character requirement of social choice functioncan be constructed to truthful trust mechanisms. Taking the VCG trust mechanism asexample, we explain in details how to construct the incentive compatible trust mecha-nisms. Moreover, we study how the trust level and recommendation weight in?uencethe recommendation acquisition cost. Finally, simulation results show that the proposedtrust mechanism can deal with all kinds of positive and negative deviations.(4) For the problem of group cheating in recommendation reporting, we proposed agroup strategy-proof mechanism from the perspective of cooperative game. The mech-anism considers the risk cost value relating to the trust value as a declared compensationand treats the solution of supermodular compensation payment game as paid compen-sation. First, we proposed two kinds of collective compensation functions in which oneis relates to the reported cost and one is not and prove the supermodularity of these twofunctions and the property of supermodular compensation payment game. Based on theselection game, we construct the corresponding indirect and direct mechanism whichis called Claiming Compensation Mechanism and prove that this mechanism is groupstrategy-proof. Furthermore, we take the Shapley value as the actual compensationpayment scheme and prove that it satisfies the property called cross-monotonicity. Analgorithm in generating the outcome coalition is given. Simulation results show that forthe collective cost function which is related to the cost value, although it is supermodu-lar, the corresponding mechanism is not strategy-proof. By adopting the collective costfunction which is not related to the cost value, the Claiming Compensation Mechanismis group strategy-proof. Finally, we summarize and demonstrate all possible results of the lying reporter.