Research On Fuzzy Comprehensive Trust Model For P2P Network

Trust Mechanism is an indispensable security policy for P2P network while trust model is a necessary part of trust merchanism. With the spread application of P2P technology, such as e-commerce, file-sharing etc, trust model is becoming more and more important. Due to traditional models’insufficient consideration in differencese of nodes’behavior when aggregating peer trust value, studying new trust model has important practical significance. After analyzing on some classical trust model at home and abroad, combining with the diversity and dynamic characteristic of peer in P2P network, this dissertation put forward a kind of trust model with the ability of dynamic adaptation. The main contents of our study include:(1) Comprehensive multi-factor trust model based on fuzzy synthetic evaluation theory. According to the problem in traditional trust model, such as rough granularity of trust model, less factors when computing local trust and insufficient consideration in peer trust behavior when aggregating peer trust etc, we propose an improved FSEM-Eigen Trust algorithm. In this method, the computing of peer trust was extended to multi-factor from single factor, and then comprehensive evaluation value was obtained through the weighted sum of each evaluation index. Finally, the efficiency of this algorithm was proved by simulating file downloding respectively at the scene of cheating in small transaction while honest in large one, and fluctuated cheating.(2) Dynamic trust model based on time attenuation factor and fluctuation punitive measure. According to the insufficient consideration in the dynamic characteric when aggregating peer trust, Based on the proposed Comprehensive multi-factor trust model, we furtherly put forward a dynamic fuzzy trust model -- DFCTrust. In this method, time attenuation factor and fluctuation punitive measure were added into comprehensive multi-factor evaluation. So, it can well avoid the failure caused by trading with inactive peer and efficiently resist attacks from periodic fluctuation cheating peers. Simulation results show that DFCTrust has good dynamic adaption ability, can efficiently deal with malicious peers regarding strategic altering behavior, and then improve the success ratio largely.(3) Dynamic fuzzy trust model based on doubly weighted average. According to the common problem existing in the current trust model -- weight setting, we proposed a dynamic fuzzy trust model based on doubly weighted average -- DWATrust. In this method, a multi-dimensional time series was applied to record the evaluations on peers in some past time segments with different indexes, then, the different weight for each evaluation index and each time segment were computed based on driving principle difference. Finally the comprehensive trust value can be obtained through data dimension reduction and twice weighted sum. Hence, the peer trust can be objectively computed from the aspect of transaction context, peer context, and the change of time. Simulation results show that DWATrust can well recognize the malicious peers such as periodic fluctuation cheating etc, then improve the transaction success ratio in P2P network largely.