Machine Learning-based Intrusion Detection for Position Falsification Attack in the Internet of Vehicles

Intelligent transportation system (ITS) is a promising technology to enhance driving safety and efficiency within smart cities. It involves public transportation management, infrastructure control and road safety. Its main purpose is to avoid risks and accidents, reduce traffic congestion and ensure safety for road users. Vehicular ad hoc networks (VANET) are core components of ITS where wireless communications between vehicles, as well as between vehicles and infrastructure, are possible to allow exchanging road, traffic or infotainment information. VANET is vulnerable to several security attacks that may compromise the driver’s safety. Using misbehavior detection approaches and information analysis demonstrated promising results in securing VANET. In this context, Machine Learning techniques proved their efficiency in detecting attacks and misbehavior, especially zero-day attacks. The goal of this chapter is twofold. First, we intend to analyze the security issue in VANET by reviewing the most important vulnerabilities and proposed countermeasures. In the second part, we introduce a comprehensive Machine Learning framework to design a VANET IDS. We used the framework to evaluate the performances of several Machine Learning techniques to detect position attacks using the VeReMi security dataset. Experimental results prove that KNN, Decision Tree and Random Forest outperform Logistic Regression, SVM and Gaussian Naïve Bayes in terms of Accuracy, F-measure, Precision and Recall. 

Keywords: Internet of vehicles, Intrusion detection, Machine learning, Position attack, Vanet security.