Study On Trust Model And Incentive Mechanism Unstructured P2P Network

P2P network technology promises a vast increase in data sharing and distributed computing applications. However, some potential risks and threats appear due to openness, anonymity and the loose-coupling relationship between peers in P2P networks, which include disseminating illegal files randomly, and abusing network resources, etc. Therefore, there is a complete lack of trust relationship between peers, and a big development for P2P network is restricted.Building trust between peers in P2P networks can motivate peers to cooperate and share resources, and improve the performance of P2P networks. Reputation-based trust management mechanism is identified as a viable solution to set up trust, encourage fairness and enhance security. But trust management model itself is vulnerable to misbehaviors. This dissertation studies the trust management model and its security problems, and the incentive menchanism is also taken into account in order to promote good development of network. We make contributions as follows:(1) Most existing studies on trust management model are limited to structured P2P networks. It has been a hot problem recently how to acquire every peer's global trust and manage them efficiently in Unstructured Peer-to-Peer Networks. A new trust management model SuperpathTrust based on trust path is proposed. It can collect feedbacks from peers for one peer, and aggregate them in the peer's neighbors, and then get global trust value by trust path computing. Thus, it solves the difficult problems of aggregation and storage for trust information. Furthermore, simulation results and performance analysis demonstrate that accuracy of trust value is high, and the overhead of trust value computing is low. Comparing with EigenTrust, the model can identify the malicious peer more easily, and decrease the probability of malicious transactions.(2) Calculating global trust value for every peer is comprehensive and convergent. But this kind of trust model can't suit highly dynamic and personalized trust environment. How to ask only part trust information of all the peers to obtain accurate reputation value and withstand malicious activity is a great challenge. A novel trust model based on parameter-estimation techniques named P-Trust to cope with misbehavior in unstructured P2P networks is proposed. In our model, each peer maintains direct trust information of others. The reputation system makes one peer evaluate other peers by the self-experiential and second-hand information. Two parameters, expectation trust value and uncertainty degree value of every trust value are considered as the metric trustworthiness. Meanwhile, incentive and punishment mechanism is taken to motivate peers to provide true information. We test and analyze the performance of P-Trust against misbehavior and the implementation costs. As a result, P-Trust can get higher ratio of successful download even under malicious collusive attacks than other models.(3) We develop a trust based network topology adaptation model for supporting dynamic trust relationship. Each peer can examine the trust relationship of its neighbors and update the neighbor list along with probability choice policy. Network topology is optimized and the successful download ratio is increasing continually. Through theoretical analysis of the model, we verify that our model is effective and robust to the malicious activities in the network.(4) In unstructured P2P networks, polluted files are disseminated by active or passive reasons, which results in the availability of network peers decreasing. A redemption mechanism based on reputation model is discussed to improve the network performance. In the trust model, every peer makes the trust evaluation of its neighbor peers locally by direct observation and use of recommend information available. In our approach, every peer maintains a trust rating for each neighbor peer. Meanwhile, trust fading and redemption are brought about by update and re-establishment of trust to show robustness. Experiment analyses indicate the trust model can exclude the malicious peers from the network. In addition, the trust redemption mechanism with trust model can save the non-malicious peers with misbehavior effectively, and enhance the network usability.