One Versus All Binary Tree Method to Classify Misbehaviors in Imbalanced VeReMi Dataset

Abstract

Nowadays, transportation networks depend heavily on the technology known as vehicular ad hoc networks (VANETs). VANETs enhance traffic control and road safety while also enabling vehicle-to-vehicle communication using basic safety messages (BSM), which are susceptible to different kinds of attacks. This study focuses on techniques for detecting and classifying misbehavior in VANETs while dealing with unbalanced data. In order to ensure equal treatment of minority and majority categories, we provide a novel method called One vs. All Binary Tree (OVA-BT). This approach separates binary classifiers for each kind of misbehavior and provides specific assessment metrics for each kind of misbehavior. We evaluate our experiment using five-fold cross-validation with six individual models of ML and an ensemble classifier. The findings demonstrated that the use of OVA-BT enhances the classification accuracy when compared to a traditional single multi-class model and that the classifier ensemble’s classification performance is greater than the best individual model on the testing set.

Keywords

• One versus all binary tree (OVA-BT) strategy
• imbalanced data
• VeReMi extension dataset
• VANETs
• misbehavior detection