Collusion detection in feedback-based reputation systems via temporal and social analysis

Nowadays, with the expansion of virtual spaces, interactions between anonymous users have also increased. These interactions can provide new opportunities and threats for users of this virtual spaces. One way to reduce these threats in such spaces is to use feedback-based reputation systems to estimate the trustworthiness of users. However, the feedback-based reputation systems are known to be particularly vulnerable to some kinds of attacks like collusion attacks between malicious user IDs. In this paper, a new method based on temporal and social analysis is proposed to deal with this kind of attack. In the proposed method, temporal analysis is used to detect users under attack as well as suspicious intervals in which the attacks have occurred with high probability; besides, social analysis can be used to distinguish between honest and malicious user IDs. The method used in the social analysis can be seen as the novelty of this paper whose main idea is the use of a small gap among the social graph of honest and malicious users to make a distinction between users in suspicious intervals. The experimental results proved that the proposed method has a better performance in distinguishing between honest and malicious users compared to earlier methods. This improvement in the performance can play a significant role in restoring and modifying users' reputations.