A Maximum Entropy Based Multi-dimensional Trust Model

To enhance security in distribute networks, it is necessary to evaluate trustworthiness of service and help selecting a proper supplier. Many probabilistic trust models have been developed, such as Beta distribution for single dimensional trust model, Dirichlet distribution for multi-dimensional trust model. But most of them are based on certain probability distribution, which are not always accurate in the practical application and will affect the precision of trust evaluation. In this paper a novel probabilistic model is proposed to compute trust value under the environment that agents exchange and combine trust information about multiple dimensional contracts. In our trust model, trust is evaluated with graded levels and trust value is computed from the multi-dimensional outcomes of the interaction among services by the joint entropy and fuzzy set principles. Finally an example is given to illustrate our multi-dimensional trust model and some related analysis has been done on its parameters. This model could effectively evaluate a supplier in multi-dimensional index with low uncertainty, provide an intuitionistic fuzzy result and could be easily satisfied different and complicated environment requirements.