ANALYSIS OF MACHINE LEARNING CLASSIFIERS TO DETECT MALICIOUS NODE IN VEHICULAR CLOUD COMPUTING

Abstract

VANET or Vehicular networks are created using the principles of MANETS and are used by intelligent transport systems to offer efficient communication between the domains of vehicles. Increasing the number of vehicles requires communication between vehicles to be fast and secure, where cloud computing with VANET is more prominent. To provide a secure VANET communication environment, this paper proposes a malicious or hacked vehicle identification system. Malicious vehicles are identified using four steps. The first step uses a clustering algorithm for similar group vehicles. In the Second step, cluster heads are identified and elected. In the next step, Multiple Point Relays are selected. Finally, classifiers are used to identify hacked vehicles. However, the existing system performance degrades as soon as the number of vehicles increases, resulting in increased cost during Cluster head election, inability to produce stable clusters, and the need for accurate and fast classification in high traffic scenarios. This work improves clustering algorithms and examines several classification algorithms to solve these issues. The classifiers analyzed are Decision Tree (DT), Support Vector Machine (SVM), K-Nearest Neighbour (KNN) and Naïve-Bayes (NB). A Hybrid classifier that combines SVM and KNN classifiers is also analyzed for its effectiveness to detect malicious vehicles. From the experimental results, it could be observed that the detection accuracy is high while using the hybrid classifier.