The Latest Developments and Future Perspectives of Artificial Intelligence Systems for In-Vehicle Communication Intrusion Detection

Abstract

CAN (Controller Area Network) is a message-based protocol that achieves communication by the exchange of packets of data between devices on the network. The protocol is widely used for in-vehicle communication (IVC) in the automotive industry as it is designed to be robust, able to handle a high rate of data transfer and tolerant of the electrical noise. However, the original CAN implementation lacks built in security mechanisms which makes it vulnerable to intrusion attacks that can be detrimental to the driver or the system itself. With the very rapid evolution of artificial intelligence (AI) various intrusion detection systems (IDS) have been developed to tackle the problem of detecting these attacks. This article aims to get a better insight of the current trend of development by surveying the latest approaches in the field of AI based IDSs. A comparison of various known attacks, detection techniques on available benchmark datasets, and a few advanced improvements of current security implementations and limitations are emphasized.