Bayesian based confidence model for trust inference in MANETs

Trust is multi-faceted, even in the same context. Nodes need to develop differentiated trust in different aspects of other nodes´ behaviours. The node´s needs vary with situation. Thus we have introduced a service based scheme for computation of trust which takes into consideration the security requirement of a node as criteria for choosing the appropriate trust computation scheme. It can either choose to use direct trust, indirect trust or to form a trust network. In this paper, we propose a modified Bayesian based confidence model that gives an explicit probabilistic interpretation of trust for adhoc networks. Then we describe trust inference algorithm, Modified SUNNY that uses probabilistic sampling to infer the trust of a node based on the highest confidence estimation. The trust network formation not only ensures the trustworthiness of the sink node but also bestows a reliable path for the routing protocol based on confidence and trust conflation. Thus we believe that our approach provides a flexible and an accurate evaluation of trust.