Trust evaluation initialization using contextual information

The majority of existing trust and reputation models consider two types of knowledge in estimating the trustworthiness of a trustee in an interaction: personal direct experiences and recommendations from third parties. However, previous direct and recommended evidence is not available for new users. In addition, a new user joins the system with a neutral reputation value in most systems and must participate in interactions with others in order to raise its reputation score. Users usually tend to interact with high reputable ones; therefore, the chance of new-comers being selected for interaction is generally rare. As a result, it is hard for a new user to raise his or her reputation score. Furthermore, shortlived users preclude the others from gaining the necessary experiences to make an accurate evaluation. Even long-lived users might leave the system and rejoin with a new identity to lose their bad reputation and start with a neutral score. Hence, effective initialization mechanism is needed to avoid such problems in trust and reputation systems. We propose to use contextual information for bootstrapping the reputation value. We use the Maximum Likelihood Estimation method for trust initialization of probabilistic trust models. We show its implementation and effectiveness for a particular model called 'Beta reputation model' through simulations.