Trust Computing And Reputatioin Managemnet In P2P Environments

Based on fully uncentralized and self-organized peer-to-peer networks, P2P (Peer-to-Peer) system makes use of network terminal's computational ability, storage ability and network bandwidth to realize broad and powerful application. However the open feature of P2P system also invites the spread of the malfunctioning data. Additional trust model and reputation systems are introduced to guarantee the data authenticity. One challenge in the design of reputation systems is how to store and spread trust value securely and efficiently.First, we analyze the inherent limitation of reputation systems based on DHT (Distributed Hash Table) networks and reach the conclusion that structured P2P systems'poor scalability makes it hard to adapt scalable and large P2P systems. Second, we analyze the features of reputation systems based on unstructured P2P networks and design a reputation system based on hierarchical reputation agents. Reputation systems based on unstructured P2P networks adopt flooding based polling mechanisms, which need to inquire into every node in the systems and cost high traffic overhead in each trade. The polling mechanisms create heavy traffic, do not guarantee voter anonymity, and make it hard for peers to filter out the fake trust values. However, our system based on hierarchical reputation agents solves the flooding problem and anonymity problem and reputation agent community is responsible for distributed storage and management of the trust values. Nodes only need sending request messages to a small group of trusted reputation agents to obtain trust values in each trade. In this way, traffic overhead can be reduced as little as possible. The reputation system based on hierarchical reputation agent makes use of public key system and onion routing to guarantee data authenticity and voter anonymity, and also makes it easier for peers to filter out fake trust values.Meanwhile, for improving our reputation system, we design a resource discovery algorithm based on trust. According to the changes of trust values, it updates the list of neighbor node and alters the topological structure of networks to reduce the time for resource finding and the traffic overhead.Finally, as the nodes in our reputation system can choose their trust model by themselves, we improve and design a general trust model which has a basic formula that combines several parameters in trade process. Reputation agents and normal nodes can design their trust computation algorithm according their interest and requirement.