Abstract: Flooding based broadcasting is the widely used mechanism in many current large scale P2P systems which usually lead to serious communication cost problem Based on the observation of rumor propagation procedure in social network, a probabilistic model of message forwarding is presented, by which the probability of message forwarding will decrease rapidly and peers usually stop forwarding when they receive enough replicated messages Mathematical analyses and preliminary simulation show that this approach could sharply optimize the communication cost in current P2P systems based on flooding mechanism, especially in those systems with clustering characteristics