Performance analysis of attack-resistant trust computation schemes in multi-hop vehicular ad-hoc networks
With the development of the autopilot technology, the autonomous vehicles mainly rely on the in-car cameras to monitor the surrounding environments. A wider view will make the autonomous vehicles safer while driving. In order to extend the view of the autonomous vehicles based on messages sharing with others, security issues should be considered. The traffic condition warning application of the intelligent transportation systems (ITS) in the vehicular ad-hoc network (VANET) could alert the road conditions ahead and assist drivers to avoid the potential dangers on the highway. An event on the traffic will be detected and reported by the vehicles passed by. The rear vehicles derive the evidence from the reports and combine them through decision logic mechanisms. The decision logic classified as the data centric trust management scheme is applied for mitigating the misleading of the untrusted reports and assisting the vehicle drivers to make correct decisions. The reputation parameter may further improve the accuracy of the decision logic mechanism by evaluating the trust of a single report. However, due to the large-scale and ephemeral properties of the vehicular networks, a secure trust computation mechanism is required for the trust assessment of the reports forwarded over multiple hops. We adopt the conventional trust computation mechanisms into the vehicular scenarios and simulate the attack-resistant performance. We mainly focus on the bad-mouthing attack and randomize the behaviors of the honest and malicious nodes based on this attack. The directed trust graph for the trust computation is generated based on the network topology of vehicles on highway. The simulations consist of the influences of the distinguishable ability of the honest nodes, the positions of the attackers in the directed trust graph, and the number of attackers in the network. Furthermore, we study the performance of the trust computation mechanisms and derive useful insights. Based on the insights, we further adjust the trust computation mechanisms and conclude their strengths and limitations. Moreover, we derive the applicable conditions for different trust computation mechanisms.