A Novel Reputation Computing Model for Mobile Agent-Based E-Commerce Systems

Mobile Agent based e-commerce systems are increasingly drawing more and more attention in recent years. However, there exist some transaction risks while enabling agents make purchase decisions and exploit information to other unknown agents in the virtual markets. Trust and reputation are widely introduced to mitigate this risk by deriving the trustworthiness of certain agent from his transaction history. Despite existing of some proposed reputation-based trust models addressing the above issue, most of them can not readily be used since there are many unforseen changes in the electronic markets. To this end, this paper proposes a novel reputation computing model that integrates a direct reputation and a recommended reputation. Specially, we present a three-factor method to evaluate the direct repu tation from personal self-experience, and adopt the vector similarity to evaluate the recommendation credibility that can effectively detect the dishonest recommendations. In addition, we amend the short term reputation and penalty factor metric to make our mechanism effective in detecting malicious agents with strategic behavior. Our experiments show that the model is highly dependable and effective.