Research Of Trust Management Mechanism In P2P File-sharing System

With the development of P2 P network, P2 P file-sharing system has been applied more and more widely. The attacks including free-riding behavior and collusion attack against P2 P file-sharing system are also increasing. These behaviors may be on purpose or not. But they actually affect the file-sharing rate of P2 P file-sharing system seriously and the download experience of users.In this paper, I proposed two trust management models to solve the issues as the waste of upstream bandwidth in P2 P file-sharing system, unfair opportunities of peers to download the resources and the reduction of file-sharing rate. By replacing the tit-for-tat mechanism by trust management, these models can defense the free-riding behavior and collusion attack, so that can improve the file-sharing rate and reduce the unfair opportunities of downloading.The first trust management model proposed in this paper is mainly to defense the free-riding behavior. This model can make each peer store the trust value of others in local without center server. This model uses both service trust and recommended trust to evaluate the performance of the behaviors of peers in P2 P network. The service trust represents the trust level of each other according to the history behaviors after transmitting a certain amount of resources. The recommended trust represents the trust value of a strange peer which can be aggregated from the neighbor peers or the neighbor peers of neighbor peers. In other words, the service trust is based on the history experience and the recommended trust is based on the recommendation of other peers. Peers can have trust values stored in other peers after they join the P2 P network. And the peers with high trust values can have a priority to get the resources. This model can also promote the fairness when peers get resources.The second trust management model proposed in this paper can defense the collusion attacks and other attacks which try to cheat other peers. This model is based on the trust value and the recommended credible value to defense the attacks. The peers which cheat others can be recognized after the iteration of the trust value and the recommended credible value. If a peer has a low recommended credible value, others may ignore the recommendation from this peer. In this way, the malicious peers cannot cheat others and cannot do harm to the system anymore. The normal peers can have a fair opportunity to download the resources.These two trust management models proposed in this paper is verified through Peer Sim: A Peer-to-Peer Simulator. The results show that these two models can defense the free-riding behaviors and collusion attacks or self-collusion attacks, can reduce the influence of the malicious peers, and can also improve the file-sharing rate.