Recommendation System to Identify Collusive Users in Online Auctions Using the Pollution Diffusion Method

Abstract

Reputation systems provide mechanisms for differentiating between honest and dishonest participants in ecommerce environments. Several reputation systems are deployed in practical electronic marketplaces. However, a considerable challenge is that malicious groups may unlawfully increase their reputation through deceitful manipulation of purchase feedback. Therefore, reputation systems must be robust and able to detect collusion, deception, and strategic manipulation. This study proposes a recommendation system to identify collusive users, inspired by observing the spread of pollution. After identifying malicious users, the suspects are likely to be identified using the proposed scheme, which efficiently reduces the relationship network size by removing relationships with positive ratings. According to the simulation results, pollution diffusion took 16 ms in a simulation involving 5000 users.