Vehicular Ad Hoc Networks (VANETs) are one of the key components for intelligent transportation systems but are still highly vulnerable to dangerous security abuses. Beyond conventional intelligent transportation systems, such secure vehicular communication is also highly relevant to smart mining environments, where autonomous haul trucks, connected service vehicles, and roadside infrastructure must exchange safety-critical messages reliably under adversarial conditions. In a blackhole attack, malicious nodes drop data packets. In this paper, we propose a new intrusion detection model based on multi-metric trust evaluation with the Random Forest (RF) machine learning algorithm. NS-3 simulation is used to generate a realistic dataset that contains five key performance indicators (KPIs) (Packet Delivery Ratio [PDR], End-to-End [E2E] Delay, Jitter, Collision Rate [CR], and Route Stability [RS]). In this model, finally RF classifier is used to identify and then quarantine malicious nodes dynamically. The experimental results show that the proposed approach significantly outperforms baseline AODV routing protocol and a recently proposed trust-based scheme. The model can maintain the PDR greater than 92% and average delay less than 42 ms, with a detection accuracy of greater than 96.3%, under the scenario with more than 30% malicious nodes as well as low false-positive rate (<2.5%). These results confirm that a multi-dimensional trust assessment with ensemble machine learning technique yields a strong and scalable solution to improve the security, safety, and reliability of VANETs.
Loading....