Research On Trust Mechanism For P2P Networks

The characters of P2P network, for example, self-organization, decentralized architecture and anonymity, result in a lot of security problems, such as, virus spreading, free-riding and cheating, etc, which cause lack of trust among peers, and the stability and robustness of network are poor. Trust relation among P2P networks is similar as the trust relation in human social networks. The trust relationship in society is introduced into P2P networks, which could make a full assessment of the peers'performance and to select the best trading partner. The P2P trust mechanism affects directly the capability of the whole system. Therefore, it's very important to build a P2P trust model which is efficient and reliable.This paper pays more attention to the efficiency, preventing the behaviors of false assessments (collusion and defamation) and strategic malicious peers in P2P trust mechanism. Based on the previous research, we build a P2P trust model based on interest group and anomaly assessment detection. In this model, we propose the direct trust and recommended trust between two interest groups, and give the formulas of calculating their weights in overall trust. These weights could indicate the peers'recent performance, which has good inhibition ability for the strategic malicious peers'behavior. To restrain the behaviors of false assessments, we find and remove the abnormal assessments in use of the abnormal assessment detection method which is based on time series. Experimental results show that the efficiency has been improved and the behaviors of selfish and malicious peers have been well controlled.Most P2P trust models pay more attention to the trust evaluation algorithm, but neglect the origin and selection of appraisal information, and they hardly give the specific methods and theoretical guidance when selecting recommendation peers. Therefore, selecting recommendation peers is low efficiency and the recommendation trust degree is low accuracy. After analyzing the peers'behaviors, we give the concept of virtual network community according to the bias and interest of transactions, and then we propose the method of selecting recommendation peers based on the probability distribution. The experiments verify that the algorithm is more effective on selecting recommendation peers and more accurate on calculating recommendation trust.