Intrusion Detection in Vehicle Controller Area Network (CAN) Bus Using Machine Learning: A Comparative Performance Study

Abstract

Electronic Control Units (ECUs) have been increasingly used in modern vehicles to control the operations of the vehicle, improve driving comfort, and safety. For the operation of the vehicle, these ECUs communicate using a Controller Area Network (CAN) protocol that has many security vulnerabilities. According to the report of Upstream 2022, more than 900 automotive cybersecurity incidents were reported in 2021 only. In addition to developing a more secure CAN protocol, intrusion detection can provide a path to mitigate cyberattacks on the vehicle. This paper proposes a machine learning-based intrusion detection system (IDS) using a Support Vector Machine (SVM), Decision Tree (DT), and K-Nearest Neighbor (KNN) and investigates the effectiveness of the IDS using multiple real-world datasets. The novelty of our developed IDS is that it has been trained and tested on multiple vehicular datasets (Kia Soul and a Chevrolet Spark) to detect and classify intrusion. Our IDS has achieved accuracy up to 99.9% with a high true positive and a low false negative rate. Finally, the comparison of our performance evaluation outcomes demonstrates that the proposed IDS outperforms the existing works in terms of its liability and efficiency to detect cyber-attacks with a minimal error rate.

A Machine Learning Model to Predict Cyberattacks in Connected and Autonomous Vehicles

“…ML techniques have shown significant potential in intrusion detection for vehicle CANs (Bari et al, 2023;Basavaraj & Tayeb, 2022;Hossain et al, 2020;Kalkan & Sahingoz, 2020;Lin et al, 2022). These methods use the capabilities of statistical models and algorithms to analyze network data and spot trends related to cyberattacks.…”

“…The result showed a FP rate of 0.078. Bari et al (2023) investigated the effectiveness of a ML-based IDS using support vector machine, decision tree, and K-nearest neighbor algorithms on real-world datasets. For the dataset provided, the models showed accuracy ranging from 93% to 99%.…”

Sybil Attacks Detection and Traceability Mechanism Based on Beacon Packets in Connected Automobile Vehicles

“…However, as CAVs grow, the cybersecurity risks they face are becoming more pronounced. The Upstream 2022 report shows that over 900 CAV cybersecurity incidents occurred in 2021 alone [3]. This grim reality has drawn the close attention of many researchers and prompted them to explore in depth the security threats faced by CAVs to propose effective prevention and response strategies [4][5][6][7][8][9][10][11].…”

Driving the Future: Design and Implementation of an Advanced Smart Car Network Simulation Platform for Education

“…CAN operates as a standardized, message-based protocol that fosters dependable and priority-driven exchanges among ECUs. Within this context, "priority-driven" signifies that communication between the ECUs on the CAN bus is orchestrated based on a predefined order of precedence [2].…”