Trust-based adversary detection in edge computing assisted vehicular networks

Low-latency requirements of vehicular networks can be met by installing mobile edge hosts that implement mobile edge computing in the roadside units (RSUs). Adversaries can, however, compromise these RSUs and use them to launch cyber attacks. In this paper, we consider an adversary that selectively drops packets or selectively corrupts packets between the RSU and passing vehicles. Such strategies would lead to a higher number of re-transmissions and thereby increase the latency of the network, which in turn impacts critical delay-sensitive applications like collision avoidance, emergency vehicle warning, etc. We propose to use trust-based detection systems to detect such an adversary. Each vehicle transmits its uplink and downlink trust values about every RSU it has interacted with. These trust values are relayed to the RSU gateway, where the decision will be made, via the next RSU encountered by a vehicle. At regular intervals, the gateway aggregates the uplink and downlink trust values obtained from multiple vehicles. It compares them against their respective thresholds to classify the RSU as benign or malicious. We also consider the presence of malicious vehicles trying to deceive the detection system by reporting false trust values. A detection mechanism is proposed to detect such vehicles. Simulation results generated using MATLAB are presented to demonstrate the performance of the proposed detection mechanisms and the impact of the adversary's parameters on the detection systems.