A novel framework for detection and prevention of denial of service attacks on autonomous vehicles using fuzzy logic

Abstract

Every year, millions of people lose their lives due to road accidents, and countless others suffer from severe injuries. Moreover, these accidents cause economic losses worldwide. Primary reasons for these accidents include over speeding, doziness, distracted driving, drugs, and psychoactive substances. Many efforts are made to automate vehicles to save losses. Autonomous vehicles are expected to revolutionize transportation by reducing the likelihood of accidents resulting from human mistakes, increasing efficiency, and reducing congestion. Like any other computer-based system, autonomous vehicles can be susceptible to attacks that could compromise their safety, security, and privacy. Denial-of-Service (DoS) attack is a cyberattack that aims to disturb a network's typical or standard functioning. A successful DoS attack could cause the autonomous vehicle to malfunction or stop functioning altogether, leading to accidents or other safety risks. To mitigate the risk of DoS attacks, protective measures that have been proposed include encryption, firewalls, and intrusion detection systems. The previous approaches to detect and prevent Denial-of-Service (DoS) attacks in autonomous vehicles are associated with imprecise and inaccurate results and high computational costs. The main goal of this research is to present a novel approach that utilizes fuzzy logic to effectively detect and mitigate Denial of Service (DoS) attacks. The proposed approach has achieved 99.4% detection and prevention rate by attaining optimal levels of security with testing error upto 0.6%. Furthermore, comprehensive execution and validation of security measures have been conducted by adopting the attack surface and rules representation. The presented approach delivers more precise and accurate results while maintaining a lower computational cost than existing methodologies. In future the proposed scheme can be used for detection of other attacks in autonomous vehicles by incorporating relevant input variables and their respective ranges.