Spreading evidence models for trust propagation and aggregation in peer-to-peer networks

摘要

Trust model plays an important role in ensuring the security of interactions in peer-to-peer (P2P) systemsrnwhere a peer’s trust evaluation depends on the interaction experience of its own and recommendation informationrnfrom other peers. However, current trust models have limitations in solving not only the issues ofrntime efficiency of direct interaction information but also the reliability and inconsistency of recommendationrninformation. In this paper, we propose a Dempster-Shafer evidence theory based trust model (ETTM) forrnP2P systems. The primary goal of ETTM is to be able to address information uncertainty and conflictingrnrecommendation problems in a reputation-based P2P environment. To make D-S theory fits into P2P applications,rnwe creatively revise the combination rules and achieve greatly improved results. To further improvernthe accuracy and performance, ETTM filters out noisy referrals if they are not compatible with most otherrnevidence. In addition, a feedback-based probabilistic searching algorithm is proposed to find the referralsrnwith improved searching success rate and lowered network traffic. Experimental results show ETTM hasrna clear advantage in aggregating recommendation information. Moreover, ETTM is more robust and canrngenerate a higher successful transaction rate than some other existing frameworks.