Trust and reputation modeling of entry and exit in evolutionary Peer-to-Peer systems

As a promising solution to file sharing and electronic commerce, Peer-to-Peer (P2P) systems attract more and more attention. Nevertheless, it's difficult to protect the participants from frauds since they have to interact with total strangers. To settle this problem, lots of trust and reputation models have been proposed. We have also designed a localized information approach for the Consumer-to-Consumer (C2C) transactions, in which we considered the scale-free referral distribution. We argue that there are two issues a trust and reputation model should consider: to distinguish malicious participants and to provide more trading chances. Few papers, however, have considered the second problem. In order to bring forward a more practical trust and reputation model, we therefore improve our previous model: in this paper we develop an algorithm to distil useful information from bias ratings instead of totally excluding them and propose new metrics, lifetime and aggregated score. Modeling a multi-agent system and running simulations on computer, we justify that our model can perform well when facing serious bias information and provide the participants with more opportunities of satisfactory transactions in an evolutionary P2P system.