Agent-based Trust Management and Prediction using D3-FRT

Reputation and trust management systems have been useful in domains that rely on the cooperation of members to function correctly and to fulfil their purposes. Despite the advent of these systems, having trusted communications remains a challenge. This is as a result of relying on the domain members for reputation information. These systems lack well analysed approaches for determining the bias of the members. A semi-distributed framework D3-FRT, which is inspired by the Dynamic Data-Driven Simulation paradigm, is presented in this paper. The framework adopts an agent-based modelling approach to make predictions about domain members. The D3-FRT framework is novel as it uses past, online and predicted data to identify misbehaving members. In this paper, the accuracy of the prediction is tested and a report on the framework's performance in different network scenarios is also presented. The results of experiments and simulations using the D3-FRT approach are discussed