Research On The Fine-grain Trust Model Based On P2P E-commerce Systems

At present, P2P e-commerce systems develops very rapidly, but because of the autonomy and dynamic of P2P network, e-commerce systems based on it usually have no trusted third parties or trust authority, the transactions are often established dynamically with peers that are unrelated and unknown to each other. For this situation, the participants have taken great risks to transact. Therefore, it is necessary to set up a trust mechanism in the P2P e-commerce systems to evaluate and measure the peer. Peers requests for transactions are able to determine whether to establish trading relationship with the other peers or not based on its trust value.On the current trust mechanism, the trust between peers often base on evaluation after the transaction. Different peers evaluate other peers on the basis of criteria that may not be the same. For example the P2P e-commerce systems, evaluate a transaction can be based on the merchandise quality, whether the goods delivers on time, whether the transaction process is successful, the vendor's service quality, merchandise domain and so on. Some peers are concerned about the vendor's trust value in one domain, while others may be more concerned about the vendor's global trust value. Therefore, it is urgent to set up a trust model that can measure the peer's trust value in any merchandise domains or any aspects (evaluation criteria), in order to make the trust computation more objectively and accurately.In order to improve the problem that the trust grain of current P2P e-commerce trust model is too rough to reflect the peers'real behavior very well. A novel Fine-Grain Trust Model (FG-trust) is presented, it can complete the trust computation that one peer in different domains and different aspects. The concept of Domain-model and Domain-trust are introduced into this model, in order to measure the trust value of one peer in any domain; the model describes the domain relationship's type that can measure the credibility of peers'Recommendation Trust when they came from different domains and presents a method to divide it; On the computing of peer's trust value, it takes various factors into account and shows the computing method for the peer's initial trust value; besides, the model use Multi-Agent structure, make it's more convenient to management and maintenance the network. Finally, the effectiveness and feasibility of this model are illustrated by the experiment.