Research On Trust Management And Search Algorithm In Unstructured P2P Network

With the recent advance in storage and networking technologies, more and more data is shared and circulated over the Internet. The traditional C/S model can not support the more and more clients. Meanwhile, the failure of single server may result in the collapse of the entire network, and becomes the bottleneck of service quality, performance and reliability. The traditional Internet approaches are facing great challenges in data sharing, cooperative working and large scale parallel computing and so on. In contrast, P2P network has abilities in distribution, self-organization, scalability, fault tolerance, AD-HOC link characteristics, low cost and anonymity and becomes the overwhelming technology for the Internet.The P2P network is an important solution for resources sharing in large-scale distributed systems. However, due to the nature of openness and autonomy, P2P network has to manage risks involved in interacting and collaborating with prior unknown and potentially malicious parties in open and dynamic network. Moreover, with the diversification of network topologies, the traditional P2P search algorithms no longer suit for the current network services. So the design of a highly safe, scalable, and effective trust management as well as its search algorithm is an important research area in P2P network.To support trust management and search algorithm with high performance and low cost, this dissertation proposes resolutions based on unstructured P2P network. The main research and contributions of this dissertation include:1,Analyze the trust in P2P network and propose a trust mechanism for unstructured P2P network to encourage resource sharing among peers and combat malicious peer behaviors. In the trust mechanism, we analyze the trust by social psychology and evaluate the trust based on Beta probability distribution. Moreover we develope aging factor, confidence and the cheat checking mechanism to prevent the malicious peers. The experiment shows that the proposed trust mechanism can autonomously deal with the biased information and identify trust belief with low overhead and better efficiency.2,Analyze the efficiency of search service and propose a Gossip algorithm with referral trust and probabilistic strategy in unstructured P2P network. The search algorithm incorporates Gossip algorithm with the referral trust to disseminate information. The main idea of the search algorithm is to use the referral trust to guide the current Gossip search. During routing, peers first calculate the forwarding probability of each neighbor according to its trust and then preferentially forward queries to the highly trusted neighbors. Since the effectiveness of any routing protocol is dependent on the P2P overlay topology, in the dissertation, a trust-based P2P topology evolution mechanism is proposed, in which peers can establish direct links to more trusted ones for facilitating interactions with them to get desired information. By establishing links to more trusted ones and removing links to less trusted ones, peers'locally adaptive behaviors can lead to globally efficient P2P networks in resource discovery. The experiment shows that the search algorithm can disseminate the query with low traffic and better efficiency.3,Analyze the information aggregation in unstructured P2P network and propose an improved reputation model based on the Dempster-Shafer theory. In the reputation model, we use the improved Dempster-Shafer theory to aggregate the reputation, in which we consider not only the reliability of the referral but also the evidences'conflict, and resolve the conflict into the uncertainty. The experiment shows that the improved reputation model solves the problems of not effectively aggregating inconsistent referrals and dealing with malicious information in unstructured P2P network better.