Machine Learning Based Approach to Detect Wormhole Attack in VANETs

VANET is one of the key enabling technologies for connected and autonomous vehicles (CAVs). In a vehicular plane of the VANETs, vehicles communicate with each other for various safety and non-safety applications. In future, the autonomous CAVs communication will not only be of type broadcast but also unicast in nature using multi-hop communication. Both broadcast and unicast multi-hop communication of VANETs are vulnerable to various types of attacks such as Denial of Service (DoS), message falsification, Sybil attack, Greyhole, Blackhole, and Wormhole attack. This paper aims to utilize the power of machine learning to detect wormhole attack in multi-hop communication of VANETs. Although various mechanisms have been proposed in the literature to detect this attack, the ML-based approach for wormhole has not been explored. To model the attack in VANET, we create a scenario of multi-hop communication using AODV routing protocol on NS3 simulator that uses the mobility traces generated by the SUMO traffic simulator. We run the simulation and collect the statistics generated using the flow monitor tool. These collected traces are preprocessed, and then k-NN and SVM are applied on this preprocessed file to make the model learn of wormhole attack. The performance of these two machine learning models is compared in terms of detection accuracy and four alarm types. Our study demonstrates that ML is a powerful tool, which can help deal with such attacks in a multi-hop communication of future generation CAVs.