Research on Cyberattack Detection of Connected Automated Vehicles Based on Support Vector Machine

Connected automated vehicles (CAVs) are facing the threat of cyberattacks while exchanging state information. Most of the existing research focuses on the communication technologies and the network security technologies, with the purpose of providing corresponding preventive solutions for the specific technical means of attack behavior, which are highly targeted but weak in universality of application. In this paper, a type of adversary model based on attack effects, namely, disturbance attack, was established. A numerical simulation platform for CAVs cyberattack experiment was built to quantitatively analyze the impacts of CAVs cyberattacks on traffic. 9 measurement indicators were constructed based on traffic safety indicators and traffic flow parameters for the characteristic impacts of various attacks. Using the cyberattack experiment simulation platform to generate attack feature datasets, A cyberattack detection model of CAVs on the basis of nonlinear Support Vector Machine was established. The model showed a sound classification performance, with a detection rate of more than 95% within 3s of the attack. The research method in this paper has a universal reference value for cyberattacks detection of CAVs. The results help better understand the effects of CAV-related cyberattacks and shed light on proactive defense against such attacks.