Control Theory Based Rating Recommendation for Reputation Systems

Reputation system is used in online markets as a trust-enforcing, deterrent, and incentive mechanism. Rating filtering mechanism is necessary in a reputation system to eliminate the impact of unfairly positive or unfairly negative ratings. The assumption that raters with low reputation are likely to give unfair ratings is not always reasonable. This paper proposes and evaluates a control theory based rating recommendation updating algorithm, which is helpful to exclude unfair ratings from reputation systems. We use recommendation trust to indicate the credibility for one´s rating. Since reputation is used as a social control mechanism, we apply PID (proportional integral derivative) control algorithm into the recommendation trust updating, which uses the rating deviation as the controller input. The experiments demonstrate that this approach can verify malicious peers quickly, especially those collusive ones. Although some improvement of reputation aggregation is needed, a reputation system with such a recommendation trust updating algorithm is feasible and effective for distributed heterogeneous systems