Bayesian-based Peer-to-peer Trust Model Research

P2P network has presented great advantage in the areas of distributed storage,share of computing capability, broadcasting in P2P application layer, etc. However,the characteristics of openness, anonymity and loose-coupling relationship betweenpeers in P2P network lead to a complete lack of responsibility for a peer can bearbitrarily walk illegal content, abuse of cyber source, which seriously restrict thefurther development of P2P.Trust is the reflections of the comprehensive evaluation from one subject toanother, the trust in P2P network is the powerful guarantee to node cooperation,encourage nodes to share resources, which directly affects the overall performance ofthe P2P system. This thesis mainly studies the trust model based on Bayesian theory inP2P environment, the main work is as follows:(1) We use the Beta distribution which is a branch of Bayesian theory to computetrust degree and propose a Beta trust model in unstructured P2P environment; theintegrated trust model we designed uses adaptive algorithm to construct unstructurednetwork, including the trust computing, storage and update.(2) We use the Dirichlet distribution which is a branch of Bayesian theory tocompute trust degree and propose a Dirichlet double trust model in structured P2Penvironment. We make the trust into service trust and feedback trust to measure thedifferent behavior of node, this method improves the accuracy of computing trust.(3) We propose some protection mechanisms against the P2P security problemsuch as Free Riding, Slander, Collusion, Oscillation attack and Strategy attack. Thesemechanisms based on the unique properties of malicious behavior. Simulation resultsshow that compared with other models our model show better performance againstthese attacks.