Cybersecurity in Automotive: An Intrusion Detection System in Connected Vehicles

• Main Authors: Francesco Pascale, Ennio Andrea Adinolfi, Simone Coppola, Emanuele Santonicola
• Format: Article
• Language: English
• Published: MDPI AG 2021-07-01
• Series: Electronics
• Subjects: cybersecurity; automotive; Bayesian network; intrusion detection system; CAN-bus; Internet of Things
• Online Access: https://www.mdpi.com/2079-9292/10/15/1765

Today’s modern vehicles are connected to a network and are considered smart objects of IoT, thanks to the capability to send and receive data from the network. One of the greatest challenges in the automotive sector is to make the vehicle secure and reliable. In fact, there are more connected instruments on a vehicle, such as the infotainment system and/or data interchange systems. Indeed, with the advent of new paradigms, such as Smart City and Smart Road, the vision of Internet of Things has evolved substantially. Today, we talk about the V2X systems in which the vehicle is strongly connected with the rest of the world. In this scenario, the main aim of all connected vehicles vendors is to provide a secure system to guarantee the safety of the drive and persons against a possible cyber-attack. So, in this paper, an embedded Intrusion Detection System (IDS) for the automotive sector is introduced. It works by adopting a two-step algorithm that provides detection of a possible cyber-attack. In the first step, the methodology provides a filter of all the messages on the Controller Area Network (CAN-Bus) thanks to the use of a spatial and temporal analysis; if a set of messages are possibly malicious, these are analyzed by a Bayesian network, which gives the probability that a given event can be classified as an attack. To evaluate the efficiency and effectiveness of our method, an experimental campaign was conducted to evaluate them, according to the classic evaluation parameters for a test’s accuracy. These results were compared with a common data set on cyber-attacks present in the literature. The first experimental results, obtained in a test scenario, seem to be interesting. The results show that our method has good correspondence in the presence of the most common cyber-attacks (DDoS, Fuzzy, Impersonating), obtaining a good score relative to the classic evaluation parameters for a test’s accuracy. These results have decreased performance when we test the system on a Free State Attack.