A Preventing Fraud Trust Model in P2P Networks

The existing trust models are not able to punish the concussive fraud in direct trust value and recommended trust value effectively. In the early stages of P2P networks, if some kinds of frauds are frequent in the meantime, the ability of trust model to prevent all the frauds is particularly important. To approve the trust model, first, referring to the idea of congestion control in computer network, "Additive Increase / Multiplicative Decrease", we put forward an adaptive punishment parameter in direct trust. Second, learning from intelligent algorithm to jump out of local optimum by random factors, we introduce a credibility parameter of trust value. And last, we import a punishment parameter in recommended trust, which is based on recommendation credibility. Simulation and analysis show that the approved trust model increases the success rate by 30% and advances the convergence time about 150 periods in the early stages of P2P network trading system.