As peer-to-peer (P2P) applications have achieved enormous success, the security of P2P network is getting more and more concerns. Particularly, the trust problem has become bottleneck because of the open and anonymous nature of P2P network. Trust model provides a security mechanism for P2P network and is proven to play an important role in improving the robustness of P2P network. However, colluding peers pose a serious threat to the security of trust model. For example, colluding peers may subvert the trust model by spreading plenty of false evaluation. Most existing trust models in P2P network are known as completely green hand in handling colluding peers. Some of them pose measures to make colluding attack difficult, but they are far from satisfying mainly because they cannot identify the colluding peers in P2P network. To address the problem, a fuzzy detector of collusion (FDC) is proposed to detect the existence of colluding peers by measuring similarity of peer's behavior. We show how to deploy FDC in distributed P2P network. The simulation results presented show that FDC is considerably effective in detecting and thwarting collusion attack and FDC can help the trust system in P2P network resist threats from colluding attackers.