Machine Learning-Based Architecture for DDoS Detection in VANETs System

Abstract

With the fast and huge vehicular communication systems currently being developed, there is a clear and urgent need for advanced security. To determine whether a vehicle has been attacked, we developed a tool called the misbehavior detection system (MDS), which is intended to help the vehicle take action and minimize any potential harm from attackers. One of the most dangerous forms of attack that threaten vehicular communication systems are distributed denial of service (DDoS) attacks, and increasing the security of VANETs against such attacks is a topic that a large number of researchers are now considering, were to provide highly effective security capabilities, machine learning (ML) techniques were applied. NSL-KDD or KDD-CUP99 datasets form the basis for the greater part of the current research. Attacks on these datasets were outdated. Therefore, we used a new dataset generated by OMNeT++, Veins, and Sumo. Two different types of attacks were conducted during this simulation, and the XGBoost classifier was used to evaluate and predict MDS systems. The median F1-score for this XGBoost classifier was 99%, which represented a clear advantage over another ML method, where we used the Synthetic Minority Oversampling Technique (SMOTE) to class balance the datasets.