As reputation mechanism has been widely accepted and adopted to enhance trust in electronic communities, how to cope with the attack and disturbance problems on reputation mechanism, such as collusion, malicious or unfair rating, becomes a key challenge. This paper extends the normal mechanism, which mainly focuses on the trustworthiness of transactions, to a double-layer reputation mechanism, by distinguishing two type reputations: capability reputation and rating reputation. Based on the double-layer reputations, we present the Rating Reputation Feedback (RRF) mechanism to confront above problems. Basic concepts, key issues, instantiated sample and the effectiveness of RRF mechanism are discussed in the paper.