An Enhanced Model for Machine Learning-Based DoS Detection in Vehicular Networks

Abstract

Vehicular communication networks should play an important role in deploying future automated and connected vehicles. Indeed, these vehicular networks could exchange information (position, speed, obstacle detection, slowing down, etc.) that could improve road safety and traffic efficiency. Therefore, it is essential to ensure the cybersecurity of these communication systems to prevent malicious entities from disrupting them. This is why, in this paper, we focus on one of the most common types of attacks in the vehicular environment: Denial-of-Service (DoS) attacks that impact the availability of services. The existing algorithms for DoS attacks detection, mainly based on Artificial Intelligence tools (Machine Learning, Deep Learning), only consider a limited number of features to build their models (position, speed). Therefore, in this paper, we quickly compare state-of-the-art approaches and introduce a new Machine Learning model considering a larger number of features and aiming at guaranteeing better performances for DoS attacks detection. We also propose an implementation and a comparative analysis of existing models to demonstrate the benefits of our approach both in terms of accuracy and F1-score.