ReFIoV: A Novel Reputation Framework for Information-Centric Vehicular Applications

In this paper, a novel reputation framework for information-centric vehicular applications leveraging on machine learning and the artificial immune system (AIS), also known as ReFIoV, is proposed. Specifically, the Bayesian learning and classification allow each node to learn as newly observed data of the behavior of other nodes become available and hence classify these nodes, meanwhile, the k-means clustering algorithm allows us to integrate recommendations from other nodes even if they behave in an unpredictable manner. The AIS is used to enhance misbehavior detection. The proposed ReFIoV can be implemented in a distributed manner as each node decides with whom to interact. It provides incentives for nodes to cache and forward others' mobile data as well as achieves robustness against false accusations and praise. The performance evaluation shows that ReFIoV outperforms state-of-the-art reputation systems for the metrics considered. That is, it presents a very low number of misbehaving nodes incorrectly classified in comparison with another reputation scheme. The proposed AIS mechanism presents a low overhead. The incorporation of recommendations enabled the framework to reduce even further detection time.