GSBDTRUST: GROUP STRUCTURE-BASED DYNAMIC TRUST MODEL FOR PEER-TO-PEER NETWORKS

Abstract

The trust model is one of the most important technologies of Peer-to-Peer (P2P) security. The popular EigenTrust reputation system can well resist the attack of malicious peers, which aggregates the extensive view of all other peers in the system to compute the global trust values for the peers, and relies on a set of pre-trusted peers heavily to ensure that the malicious peers do not subvert the system. Nevertheless, the EigenTrust model, which designates early participants or creators of a P2P network as pre-trusted peers statically, poses a high security risk, especially for the targeted attacks. To solve this problem, it is an inevitable trend to select pre-trusted peers dynamically according to the interaction between peers. And thus, this paper selects pre-trusted peers dynamically based on the time-interactivity of peers in the network, which can avoid a series of problems caused by statically specified pre-trusted peers. Meanwhile, the peers can not be effectively organized together for network models with community structure, which increases the unnecessary interaction between peers greatly, and aggravates the load rate of the system heavily. In order to better organize peers together in the distributed network, this paper proposes a P2P network trust model based on group structure based on the interaction characteristics of peers. To ensure the high credibility of P2P network, we choose the peers with high global trust values as pre-trusted peers in each group. The final simulation experiments show that the GSBDTrust model not only has strong robustness against malicious peer attacks with different thread models, but also has low query overhead and failure rate in resource search.