Poisonedwater: An improved approach for accurate reputation ranking in P2P networks

• Authors: Yufeng Wang;Akihiro Nakao
• Venue: Future Generation Computer Systems
• Year: 2010

It is argued that social-network based (or group-based) trust metric is effective in resisting various attacks, which evaluates groups of assertions ''in tandem'', and generally computes peers' reputation ranks according to peers' social positions in a trust graph. However, unfortunately, most group-based trust metrics are vulnerable to the attack of ''front peers'', which represents malicious colluding peers who always cooperate with others in order to increase their reputation, and then provide misinformation to promote actively malicious peers. In traditional reputation ranking algorithms, like Eigentrust and Powertrust, etc., front peers could pass most of their reputation value to malicious friends, which leads to malicious peers accruing an improperly high reputation ranking. This paper proposes an alternative social-network based reputation ranking algorithm called Poisonedwater, to infer more accurate reputation ranks then existing schemes, when facing front peers attack. Our contributions are twofold: first we design the framework of the Poisonedwater approach including the following three procedures: (1) the propagation of Poisoned Water (PW): through direct transactions or observations, several malicious users are identified, termed as the poisoned seeds, and the PW will iteratively flood from those poisoned seeds along the reverse indegree direction in the trust graph; (2) the determination of adaptive Spreading Factor (SF) from PW level: based on the logistic model, PW level will correspondingly shrink each peer's adaptive SF, which can determine how much percentage of each peer's reputation could be propagated to its neighbors, and can be regarded as indicative of the peer's recommendation ability; (3) the enhanced group-based reputation ranking algorithm with adaptive SF which seamlessly integrates peers' recommendation ability to infer the more accurate reputation ranking for each peer; second, we experimentally analyze the mathematical implication of the Poisonedwater approach, and investigate the effect of various parameters on the performance of Poisonedwater. Simulation results show that, in comparison with Eigentrust and Powertrust, Poisonedwater can significantly reduce the ranking error ratio up to 20%, when the P2P environment is relatively hostile (i.e., there exists a relatively high percentage of malicious peers and front peers).