Research On Trust Management Model In P2P Network

With the growing popularity of Peer-to-Peer technology, the issues of security in P2P networks arouse more and more scholars'attention. The resources and nodes'responsibilities are distributed at all nodes in P2P networks that it is hard to implement authentication and authorization strategy without a central Certification Authority. The lack of trust relationship between peers provides the malicious peers more opportunity to attack the network, and malicious nodes can even make use of security loopholes to promulgate viruses and malicious content. However, subjective trust management model for P2P networks can provide reliable support for trust decision-making and it is worth researching in theory and application.Currently, the majority of research in building a subjective trust management system employs the social network simulation to model the formation of trust. However, many models overlooked the subjectivity and ambiguity of the trust relationship. This paper uses fuzzy theory and method to measure and compute the trust, and designed an overall Subjective Trust Model Based on Fuzzy Theory—FSTM.The model FSTM uses the fuzzy set theory to express the fuzzy semantics when it assesses the trust value. The trust model can evaluate the performance of different aspects of the nodes. At first, the fuzzy operator Einstein is improved to define the connection and merger operator of trust Vector; then the two operators are used to compute direct trust value, local trust value between different peers and peers'global reputation; The forming of trust chain of the network is based on direct trust value which is mainly calculated by integrating historical record of the transactions with some parameters such as time attenuation coefficient and the mumber of trades; Local trust value is integrated by direct trust value and global reputation with proper weight and its result represent the degree how much a peer trust another peer; The all peers'local trust value to some peer merge into its global reputation,whose updating algorithms is comparatively complex; Many algorithm renew it by continuously iterating the trust relationship until obtaining a stable state, however, the paper presents a method which computes the transitive closure of global trust relation fuzzy matrix firstly and then updates the vector of global reputation and the method is identical with iterative methods but avoids the problem of the iteration convergence; This paper also presents a distributed data storage scheme that a peer preserves local empirical data trading with other peers locally, while preserves the comprehensive evaluation data of other peers in some peer with storage mechanism of distribution hash tables (DHTs); The method can solve the problem of propagating trust data well and can prevent tampering and collaborative fraud attacks effectively; Then the paper gives a distributed collaborative protocol and applied it to the trust management agent algorithm; At last a P2P file-sharing system is achieved to verify the validity and usability of the model, and simulation shows that the calculation of the trust updating and network message costs are relatively low;The system can improve the rate of successful barter greatly, and eliminate some secure problems such as slandering, venomous recommendation, etc.