Research On Trust Models For P2P Networks

P2P service has become the one of the most important applications in the Internet nowadays. However, the nature of openness, anonymity and loose-coupling relationship between peers in P2P networks leads to a complete lack of responsibility for the content a peer puts on the network, opening the door to abuses of these networks by malicious peers. Moreover high dynamics of peer behaviors always bring on unsteadiness of the network bandwidth and performace. Therefore, it is necessary to build trust relationship between peers in the distributed large-scale P2P networks.Trust represents an integrated evaluation of behavior and capacity of a peer from the viewpoint of another peer. Trustworthiness between peers in P2P networks can incent peers to cooperate and share resources. Current researches on trust mostly focus on setting up the reliable trust management model for P2P networks. In this dissertation, we go deep into the trust models of P2P networks. The main achievements are as follows:(1) A novel reputation and risk evaluation based trust management model is proposed. The trust model consists of two parts: reputation and risk evaluation. Reputation represents the accumulative assessment of the long-term behavior, while the risk part which is quantified using information entropy is employed to deal with the uncertainty of peer behaviors. As to the problem of security, some measures are proposed to defense against several kinds of malicious attacks. Subsequent experimental results show that, by adjusting the weights of reputation and risk value, our model is more robust on trust security and more advanced in successful transaction rate than other trust mechanisms.(2) A Recommendation Evidence based Trust Model (RETM) for P2P networks is proposed. It solves the problems of not effectively aggregating inconsistent referrals and dealing with uncertainty of information in the reputation-based P2P trust model. An enhanced D-S combination rule which is obviously superior to the original one is put forward. Before combining the evidences, RETM filters out noisy feedback information using a useful filtering algorithm. Experimental results show that, compared to some current trust models, RETM has advantages in modeling dynamic trust relationship.(3) An Attack Resistant Trust management model (ARTrust) is proposed to effectively evaluate and precisely update the credibility of a recommender in reputation based trust models. It is proved that it can efficiently distinguish reliable peers from deceptive or unreliable peers. In addition, a feedback-based probabilistic searching algorithm is presented to find the referrals. It improves the searching success rate, the average search path and lowers the network traffic.(4) A Super-peer based Trust model (SuperTrust) for P2P networks is put forward. In SuperTrust, the trust value of a peer from the viewpoint of another in a group is computed based on the reputation, while that of a super peer in the group is evaluated by global trust value. A filtering algorithm is proposed to filter fake, useless and misleading ratings. Finally, the simulation results have validated the effectiveness of the trust mechanism.