Trust Model In Multi-agent Systems

E- commerce represents the future of commerce. It has created new channels and opportunities for business. Agent as a special computer system capable of flexible autonomous action in a dynamic, unpredictable and open environment has played an important role in e-commerce. However, agent-mediated e-commerce also generates a series of issues that remain unsolved. One of them is trust. Agents interact with one another typically by making commitments to one another to carry out particular tasks (transactions) in agent-mediated e-commence systems. However, in most realistic environments, there is no guarantee that a contracted agent will actually enact its commitments. This can happen because an agent may defect to gain higher utility or because there is uncertainty about whether the task can actually be achieved. Agents are therefore unavoidably faced with significant degrees of uncertainty in making decisions. In such situations, agents have to trust each other in order to minimize the uncertainty associated with interactions in open multi-agent systems.Trust plays an important role in the success of e-commerce. Different approaches for trust management have been proposed and studied in different domains, but no one has been proved efficient and effective for agent-mediated e-commerce. In this paper we propose a computational trust model with trust network so that agents can autonomously deal with deception and identify trustworthy parties in multi-agent systems. Firstly, this paper introduces the basic concept of trust, the status and trend in trust management; analyze the art-typical trust models in multi-agent systems. Then it puts forward the computational trust model with trust network based on Ramchurn's trust model with confidence and reputation, and specifies the design ideology. At last, it compares the model and draws a conclusion.The computational trust model with trust network advances the state of the art in the following ways. First, when evaluating the trustworthiness of the target agent, our approach not only considers the confidence information (based on direct prior interactions with the target agent) and the reputation information from trust network, but also considers the reliability of the witnesses. Second, we use the Weighted Majority Algorithm (WMA) to deal with deception in multi-agent systems. Finally, our model considers the ontological property of trust, which combines trust on different aspects to calculate a more complex one.In the end, a case study of travel agency is provided to show the effectiveness of the proposed model.