Research And Improvement On EigenTrust Based On P2P Networks

In peer-to-peer networks,EigenTrust is a widely used trust evaluation model.At present,when peers in P2 P networks transact with each other,trust evaluation models are adopted as the basis for the trust among nodes to improve the security and success rate of transactions,and one of the most representative models is EigenTrust which is based on no center certification.EigenTrust calculates the trust values based on the number of satisfactory and unsatisfactory transactions.It provides the security mechanisms to inhibit dishonest services of the peer.However,there are still some problems currently.In this article,the defects of EigenTrust are analyzed and improved in order to solve these problems.The main contents are illustrated as follows:(1)The calculation method and the trading process of EigenTrust model in P2 P networks are researched and analyzed.The result shows that,first of all,since EigenTrust model only adopts the cumulative value of the difference between the number of satisfactory transactions and the number of unsatisfactory transactions as the basis for calculating local trust value,it ignores the effect of historic trust information on the current trust value calculation,as well as not fully considers the time-series consistency of transactions.Secondly,since EigenTrust model uses the same way to handle the satisfactory transactions and unsatisfactory transactions,it leads to the difficulty to distinguish between malicious and non-malicious peers,and a poor effect of the punishment for the malicious peers.Lastly,since EigenTrust model uses unified normalized method to dispose local trust value of all the nodes in P2 P network,it leads to the assimilation of heterogeneous nodes,and the calculation result of the local trust value deviating from the truth.(2)To solve the problems in the calculation method of local trust value,the calculation method of local trust value based on transaction history time factor and punishment mechanism is proposed.In order to solve the problem that Eigen Trust model does not consider the transaction time factor,this article introduces a concept of time attenuation factor,which gives larger weight coefficients to recent performance of peers and gives less weight coefficients to previous performance of peers,synthesizes time attenuation factor in the formula of trust,and updates calculation method of the local trust value.In addition,EigenTrust model doesn't have the punishment mechanism for malicious nodes,so this paper proposes an improvement with incentive/punishment.When the behaviors of peers get poor,trust values will fall quickly;when the behaviors of peers get better,trust values will rise slowly.It meets the process of establishing trust,and also reaches the effect of punishing swinging nodes.(3)In order to solve the problem of the service swing peers in P2 P networks,the paper introduces and improves an existing punishment mechanism based on the degree of service swinging,and proposes an authentication mechanism and restraint strategy for Eigen Trust model.First of all,a history array m is set for all peers in EigenTrust model to store service swing records of peers.The array is a one-dimensional vector,whose index indicates the weight of trust value and it increases in turn from left to right.Second,after the peers begin to transact and the global trust value is calculated,the service swing peers can be judged by calculating the change rate of the difference of the local trust values in two consecutive cycles.Finally,the last number of the array can be assigned value according to the change rate of trust value.The higher the change rate is,the bigger the value is.Thus the degree of service swinging is computed and synthesized into formula of the trust value.(4)PeerSim is used as the simulation experiment platform in this article,and necessary codes are added to simulate documents sharing application in P2 P network.Two main experiments are the effect experiment of motivation/punishment mechanism with time attenuation factor on the change of trust value,and the inhibition effect experiment of the degree of service swinging on dynamic behaviors of peers.The results show that,the improved model in this paper has a better punishment effect compared with traditional EigenTrust model.It enhances the binding of the model on dynamic behavior of malicious peers in P2 P network.And it has a greater ability of authenticating and restraining malicious service swing peers.It can overcome service swing attack effectively and promote the probability of honest services and achieve the purpose of maintaining transaction order in P2 P system.