Reasearch On Trust Model In Mobile P2P Network

Abstract: Traditional P2P network can exchange network resources and services directlybetween peers. P2P technology’s core idea is using distributed storage and computing power tosubstitute centralized mode. Mobile P2P network is a new era of traditional P2P network. MP2Pnetwork’s main function is promoting mobile devices’ collaboration and resource sharing throughdirect data exchange. MP2P network can make good use of limited resources on a larger scale dueto it has ability to promote resource-constrained terminals working together.Trust relationship is an important issue for MP2P network security. In MP2P network, deviceidentity, addressing and communication mode are different with traditional P2P network. Somesecurity issues such as: allusion attack, malicious slander and collusion,“free ride” behaviors areeven more serious compared to traditional P2P network. Aimed at these problems, a Proxy-basedSecurity-Feedback Trust Model (PSTM) is proposed. Certificate Feedback Rater’s identificationand qualification through security resource-selection protocol then integrate trust information.Furthermore, set global contribution and evaluation value in multi-granularity trust computationand divide mobile peer’s direct trust value into peer-oriented and resource-oriented values tomake trust data more authentic. Simulation experiments show that PSTM can reduce maliciousslander and collusion effectively. It also can restrain selfish peers’“free ride” behaviors andincrease successful cooperation rate of high-contributed peers in MP2P network.When mobile peers cannot get global credibility value and contribution value from proxyserver, this thesis proposes a trust model based on fuzzy Q-learning algorithm trust model(FQTM). FQTM model is designed to maximize mobile peers’ benefits through learning andadapting unfamiliar environments. This model applies fuzzy theory into Q-learning algorithm inthe calculation process of state transition probability. Mobile peers adjust their decisionsconstantly based on past records with the policy iteration algorithm. Then FQTM updates Q valueuntil it converges to a relatively stable statement. Experimental results show that FQTM has afaster speed than greedy strategy in Q-value convergence and it also can help mobile agents toachieve their benefits maximum.