DIRECT: A robust distributed broker framework for trust and reputation management

Reputation systems for e-cammerce uses past behaviors as a predictor of future behaviors. Reputation systems must meet the challenge on how to differentiate honest feedbacks from dishonest ones in order to reduce the system vulnerablity due to malicious attacks. To address this issue, we propose a distributed reputation and trust management broker framework, called DIRECT, for dishonesty prevention. In DIRECT, malicious attacks are monitored and detected by local and cross-broker check mechanisms using statistical distribution test technologies. In the presence of dishonest feedbacks, DIRECT effectively classifies users into the green (good) and red (bad) groups. It saves the correct reputation information provided by the green group, and reduces the impact of dishonest feedbacks from the red group. The performance study shows that the DIRECT sanity check mechanism works effectively in the presence of dishonest and malicious feedbacks.