A Trust Domain-based Multi-agent Model For Network Resource Selection

The World Wide Web has evolved at an extreme rate due to its capacity to provide an endless amount of resources to the public users. And the P2P network has developed faster than the network of C/S model for its convenience and flexibility. Hence, the user finds him lost in a pool of information, without knowing how to select resources and which resources are useful and credible. A wide range of mechanisms such as contracts and commercial laws as well as face to face meetings help reduce the likelihood of risks to the consumers in the traditional businesses. However, when doing online trading, users often have little or no prior knowledge of their potential business partner(s) and the absence of these mechanisms and face-to-face encounter make them cannot check goods before paying and cannot distinguish cheat from honest effectively. As a result, there has always produced a trust deficit in e-commerce.Trust model is the framework to set up and manage trust relationships between peers in the e-commerce transactions of the network. It presumes the peers'behaviors through evaluating the reliability of the peers in the P2P network. The purpose is to make the behavior of the peers standardized, avoid malicious transactions and maintain the security of the transactions and the network.The problems of the credibility of on-line trading and the exact selection of network resources have become the key factors of effecting the development of e-commerce. For this reason, a trust domain-based multi-agent model for network resource selection is presented. The model divides the network into numbers of trust domains. Each domain has one domain-agent and several entity-agents. The model prevents the inconsistency of information that maintained by different agents through the periodically communication between the agents. The model calculates global buyer reputation and global seller reputation for each node to distinguish the credibility of node in different transaction behaviors. In order to make users choose the best matching services to their request and give users with trusted services, the model takes the recognition value that the service provider to the service and the similarity between services into account and uses the similarity degree to amend the integration reputation value with harmonic-mean. And the model enables the consumer to receive responses from agents significantly quicker than that of traditional models, because the global reputation values of service providers and consumers are evaluated and updated dynamically after each transaction. Finally, the model is illustrated the effectiveness and feasibility according to the experiment.