Reputation reporting systems have emerged as an important risk management mechanism in online trading communities. However, the predictive value of these systems can be compromised in situations where conspiring buyers intentionally give unfair ratings to sellers or where sellers discriminate on the quality of service they provide to different buyers. This paper proposes a set of mechanisms that eliminate or significantly reduce the negative effects of such fraudulent behavior. The proposed mechanisms can be easily integrated into existing online reputation systems in order to safeguard their reliability in the presence of deceitful buyers and sellers.