Towards Misbehavior Intelligent Detection Using Guided Machine Learning in Vehicular Ad-hoc Networks (VANET)

Abstract

C-ITS (Cooperative Intelligent Transport Systems) is a new technology that aids in the reduction of traffic accidents and the enhancement of road safety. VANETs (Vehicular Ad hoc Networks) are an ITS system based on inter-vehicle communication through the transmission of basic safety messages (BSM), which are vulnerable to a variety of misbehaviors. To resolve this challenge, we developed in this paper an automated learning approach-based misbehavior detection system (MDS) to identify and categorize misbehaving messages delivered by a vehicle on VANETs using VeReMi extension database. This study examines different types of classification: in binary classification, all sorts of misbehavior were grouped into one "misbehavior" category; however, in multi-class classification for three classes, misbehavior was divided into two classes: attacks and faults. The classifier has substantial issues when learning from unbalanced data when working with multi-class issues, this gets considerably more complex. The relations between categories are no longer well-defined, and it is easy to lose effectiveness in one class while improving in another. As a result, the findings are uncovinient in the Classic Learning Approach for Multi-class Classification when classifying misbehaviors into different types of misbehaving classes. To solve this issue, we developed a novel and powerful approach named "Guided Learning Approach for Multi-class Classification" to reduce the number of classes by combining comparable misbehaviors into one. According to the results, the Random Forest classifier outperforms other classifiers.