Analysis of machine learning algorithms for DDoS attack detection in connected cars environment

Cyberattacks against the Internet of Vehicles (IoV) will continue to evolve as the industry continues to adopt new and advanced connected technologies. These advances should result in a complex ecosystem that integrates different technologies (5G, 6G, Cloud, IoT, etc.) and presents a large attack surface. Denial of service (DoS) attacks are among the most dangerous attacks against connected vehicles, where an attacker overwhelm the network with random generated messages. An effective detection of the occurrence of such attacks is a key step for any defense scheme. Machine-Learning (ML) and Deep Learning (DL) algorithms have attracted a lot of attention in the literature for DoS detection. However, there is no comprehensive comparative analysis of their efficiency in the IoV context. In this paper, we study the detection performance of several classification algorithms such as Decision Tree, Random Forest, XGBoost, AdaBoost, Logistic Regression, Support Vector Machine (SVM), Naive Bayes, and K-nearest neighbors to differentiate normal CAM messages from flooding CAM ones launched by the vehicular bot in Vehicle-to-Infrastructure (V2I) environment. The obtained results demonstrate that XGBoost, Random Forest, and SVM algorithms have a very high detection precision.