An Integrated Approach to Securing In-Vehicle CAN Bus Network Using ECU Fingerprinting and Image Classification Techniques

Abstract

This paper presents two novel techniques for securing electronic control units (ECUs) in the controller area network (CAN) bus network of autonomous vehicles (AVs). Method 1 proposes an ECU fingerprinting technique to detect malicious ECUs in the in-vehicle CAN bus network. The technique extracts unique digital signatures based on intrinsic characteristics of the ECUs to identify the ECU responsible for broadcasting counterfeit messages received on the CAN bus. The proposed work employs three machine learning (ML) algorithms, namely k-Nearest Neighbors (k-NN), Support Vector Machine (SVM), and Logistic Regression (LR), to analyze the data from seven distinct ECUs. The performance of the proposed cybersecurity framework is evaluated and compared using these algorithms. Method 2 proposes a solution for efficient ECU classification to protect against multiple threats and attacks, including hacking and spoofing attacks, that can be perilous for the vehicle and its occupants. This technique visualizes signal loss and distortion in the ECU voltage signals caused by the ECU position in the CAN bus and utilizes Euclidean distance-based image classification on the principal components. This methodology is based on the commercially feasible eigenface-based algorithm, which has found extensive application in real-time scenarios like face recognition, fingerprint recognition, image recognition in photo-based applications, and real-time object identification. By profiling the ECU using this method, we can make the signal analyzer commercially viable. In this paper, we analyze different ECU configurations in the daisy chain network to evaluate the effectiveness of our proposed method. Our approach achieves a high accuracy rate of 97.14%. The proposed techniques address the security concerns related to the CAN bus network of AVs and provide efficient and effective ways to secure ECUs against malicious activities. The use of machine learning algorithms and visualization techniques in these methods not only enhances the accuracy of ECU detection and classification but also provides a better understanding of the underlying data. These techniques can be implemented in AVs to improve the security of the CAN bus network and ensure safe and secure operation of the vehicles.