Research On Trust Models For P2P Networks

Since the P2P(Peer-to-Peer) appeared, P2 P has become a hot topic in the Internet. Compared to the traditional Client/Server model, P2 P can improve the utilization of network resources effectively. However, openness, anonymity, nodes autonomy and some other characteristics of P2 P networks make its drawbacks exposed gradually, malicious nodes transmit the virus and the unreliable documents. Meantime, resources of nodes are lack of effective protection, and there is the possibility of being theft. These security problems seriously restrict the development of P2 P networks.As the security problems in P2 P networks highlighted, it has promoted the study of trust management in P2 P networks. Trust management is an effective way to maintain the level of credibility of the whole system and encourage faithful cooperation between nodes in the network.This paper is mainly based on the research of the P2 P trust model based on trust value, the main contributions are as follows:(1) To improve the P2 P trust model’s ability of inhibiting malicious nodes, especially the increasing“free-rider” nodes in traditional trust model for P2 P network. Grouping P2 P trust model based on dual attributes is proposed. Grouping nodes in the network according to the level by using the basic fuzzy inference rule and combining trust value and contribution value, and limiting the resource access through the level of the node. Introduced time attenuation function reflecting the actual situation and penalty factor punishing malicious behavior of the node in the calculation of the comprehensive trust. According to the analysis of data, the proposed model can effectively inhibit the malicious nodes attack.(2) In the early stages of P2 P networks, if a variety of malicious attacks occur frequent in the meantime, the existing trust model’s ability of inhibiting malicious attacks is insufficient. To improve the problem, an effective preventing concussive nodes and collusive nodes trust model in P2 P networks is proposed. First of all, in order to make the adaptive punishment parameter can change as the change of concussive behavior while calculating the direct trust. Referring to the idea of congestion control in computer network, “Additive Increase / Multiplicative Decrease”, we put forward an adaptive punishment parameter in direct trust; Second, to prevent the success rate of the trading system is too low in the early stage, learning from intelligent algorithm to jump out of local optimum by random factors, we introduce a credibility parameter of trust value; And last, in order todecrease the recommend trust value of concussive nodes we import a punishment parameter in recommended trust; Simulation and analysis show that the approved trust model can effectively increase the success rate and decrease the convergence time of system.