Adaptive thresholding and neural network-based misbehavior detection for incentivizing VANET data sharing
Vehicular ad-hoc networks (VANETs) are highly dynamic traffic networks made up of vehicles and infrastructure nodes like Road Side Units (RSUs). In VANET safety applications, data sharing increases the situational awareness of nodes minimizing traffic accidents and congestions. In practice however, it is unlikely to recruit nodes to join a safety application network without providing an incentive. In this paper, we propose a system that incentivizes data sharing in VANETs using crypto tokens running on a lightweight Blockchain. Our system rewards data sharing nodes while mitigating against malicious node collusion and node misbehavior. Nodes in VANETs periodically broadcast a beacon known as the Basic Safety Message (BSM). In the proposed system, nodes create a trajectory of handshakes by digitally signing each other’s BSMs. Each node locally queues its handshakes until a certain threshold queue size is reached at which point the queue gets uploaded to the cloud infrastructure to earn reward tokens. Unlike previous works that deal with arbitrary or global thresholds, we design a scheme of setting regional, dynamic and tunable thresholds. Furthermore, we have trained a neural network model on the VeReMi dataset to help detect location falsification during node-to-node communication.
Committee: Ali Jannesari (major professor), Ying Cai, and Qi Li