A protocol based on a game-theoretic dilemma to prevent malicious coalitions in reputation systems

Abstract

In decentralized and open systems, a large number of agents interact and make collective decisions in order to share resources. As those systems are open, the presence of malicious agents needs to be considered. A way to deal with such agents in a decentralized fashion is to use reputation systems. But, as reputation systems are based on the aggregation of local trust between the agents, they are vulnerable to malicious coalitions, particularly to self-promotion based on false identities. In this paper, we propose a game-theoretic approach to prevent such manipulations. Its main feature is that honest agents use in turn a false-name manipulation to fool malicious agents and to drive them into a dilemma. We show that the best response to that dilemma in terms of mixed strategy equilibrium leads the malicious agents to give up most of their manipulations.