Data Generation and Verification for Development of DoS Attack Detection Model in V2V Communication Environment

Abstract

In recent years, autonomous vehicles have been using deep learning to recognize road conditions and make driving decisions. In addition, autonomous driving that uses only deep learning technology has limitations, so it utilizes vehicular ad-hoc network (VANET) communications. However, VANET communications contains vulnerabilities that can be exposed to cyber-attacks such as denial of service (DoS), and research is underway to defend against them. In this paper, we generate a dataset to develop a machine learning model that can detect DoS attacks in the V2V communications environment of VANETs. The dataset is generated using simulation tools, such as OMNeT++, SUMO, Veins, and INET, to reflect the attributes of V2V communications and characteristics of the attacks. In addition, the attack dataset generated is validated to see if attacks can be detected by various machine learning models. The evaluation results show that the generated dataset can detect DoS attacks with an accuracy of about 97% or higher from most of the trained machine learning models, which is useful for training intrusion detection models.