Research On Grouped Global Trust Model

P2P (Peer-to-Peer) networks has become a hotspot for research as the development of the network. It collects idle resource from terminals, such as computing, storage capacity, network bandwidth etc. improve the efficiency of massive storage and computing. Therefore, it has brilliant potentials in distributed computing, system collaboration, deep search, multimedia and many other areas. However, it also faces many challenges. Its characteristics, includes openness, anonymity, dynamicity, could bring users not only convenient service, but also security risks, such as malicious nodes' invasion, viruses, junk files spreading and so on. Therefore, it's necessary to construct an efficient trust management system for P2P.Establishing a trust model is to find a quantitative evaluation system to measure the "credibility" of nodes'. It ensures for nodes the security of computing, storage and distribution without losing the essential attributes of P2P networks. A good trust model should be able to encourage cooperation of peers, reduce costs and improve network performance. Therefore, the paper proposes a new trust model GGTM (Grouped Global Trust Model) based on above considerations.The paper mainly consists of the following parts:First, it introduce the basic concepts of P2P networks and trust model in details. It also does some research on current development of trust models and analysis the feathers of several mature trust models.Then, it gives out the design of GGTM. This model adopts a grouped architecture, which reduces system cost significantly, makes resource searching and trust evaluation more efficient, and improves the performances of network. The model calculates trust evaluation in the whole group manner and has a good defense on malicious/selfish nodes; the improvements of the dijkstra algorithm, which includes considering the recommended credibility, carefully considering the extreme trust evaluation, and updating measures per time slot, bring strong resistance to cooperative cheating; besides, the model propose a "double defense" to prevent white washing. The management of inter-group ensures the safety and effectiveness of grouped interactions.Finally, the simulation proves the convergence of the model, which avoids the iterative computing problems met by some previous models. So it shows good feasibility and adaptability for large-scale projects. The hierarchical management avoids the "single point failure" problem. We use C++ implemented simulations to collect runtime model statistics and draw graphics according to the results with matlab. After comparing the results with that of SWRTrust, we can prove the good performance and effectiveness of GGTM.