Simplified clique detection for collusion-resistant reputation management scheme in P2P networks

Abstract

Reputation management systems have been proposed to provide helpful information for reliable peer selection in an environment where honest peers coexist with malicious peers. The reputation management systems generally seek to generate an accurate assessment in the face of various factors including but not limited to potentially adversarial environments. Thus, performance of the reputation management systems mainly depends on an effectiveness of detection of various malicious attacks. In this paper, we focus on a detection of colluders who usually form a clique among various malicious attacks since only heuristic-based approaches which are still vulnerable to sophisticated colluding attacks have been proposed due to a NP-completeness of the detection of colluders forming a clique. For collusion-resistant reputation management, we introduce a simplified clique detection that can be applied to the reputation management system so that the colluders forming a clique can be easily detected. Then, we introduce a way to calculate a collusion probability-weighted reputation to reduce falsely cumulated reputation by malicious collusions. Through simulations, we show that our approach is enough to detect colluders forming a clique and shows better performance than heuristic-based approach in terms of authentic file download and reputation management.