@Trust: A Trust Model Based On Feedback-Arbitration In Structured P2P Network

With the development of P2P technology, P2P networks have become one of the most popular application systems. Because of their natures of open, anonymity and autonomy, however, nodes in networks lack trust mutually and thus resulting in obstacles to the development of P2P networks.Some studies have shown that a reliable trust management model established in P2P networks could resolve the trust problem. Based on the past evaluations from other peers for this node, the trust model could predict a node's future behavior, which could promote the establishment of trust relationship and assist other nodes to make the right decision, as well as against malicious behaviors between transactions. Now, the trust model has become a hot research topic in P2P networks. In the current studies, many researchers, from different perspectives, have been presenting varieties of trust management models for different applications in P2P networks. However, these trust models barely consider the actions of false feedback or lack ability to effectively process false feedback information, in addition, the processing granularity is too coarse in the aspect of trust decision-making.Aiming at shortcomings in feedback information aggregation and trust decision in existing trust models, a novel trust model, named @Trust, is proposed for structured P2P networks. According to the rule of majority in our society,@Trust introduces an arbitration mechanism which would arbitrate the credibility of that evaluation, based on the past evaluations from other peers for the same specific service, and then calculates the node's credibility of service, feedback and arbitration based on the arbitration result, to enhance the model against the malicious peers. What's more, this paper also proposes the concept of the credibility of specific services to improve the trust decision, a node will consider the partner's credibility and the specific service's credibility when begin a transaction. Aiming at the attack behaviors of malicious nodes, we introduce punishment mechanism into our trust model; specifically, the model applies different punishment strategies to different behaviors in order to isolates normal peers from inauthentic resources effectively. Finally, based on the structured P2P network, Chord, this paper systematically presents the implementation of @Trust, which includes the update, saving and management of the various kinds of credibility, as well as the transaction process. The simulation results show that @Trust has advantages in distinguishing out the false feedback and resisting the malicious peers, and thus improves the availability of P2P networks.