A Novel Hybrid Quantum-Classical Framework for an In-vehicle Controller Area Network Intrusion Detection

Abstract

<p>Controller area network (CAN) is susceptible to various cyberattacks due to its broadcast-based communication nature. In this study, we developed a hybrid approach for CAN intrusion detection using a classical convolutional neural network (CCNN) and a quantum restricted Boltzmann machine (quantum RBM). The CCNN is dedicated for feature extraction from CAN images generated from a vehicle’s CAN bus data, while the quantum RBM is dedicated for CAN image reconstruction for a classification-based intrusion detection. To evaluate the performance of the hybrid approach, we used a real-world CAN fuzzy attack dataset to create three separate attack datasets, where each dataset represents a unique set of features related to the vehicle. We compared the performance of our hybrid approach to a similar but classical-only approach. Our analyses showed that the hybrid approach performs better in CAN intrusion detection compared to the classical-only approach. For the three datasets considered in this study, the best models in the hybrid approach achieved 97.5%, 97%, and 98.3% intrusion detection accuracies, and 94.7%, 93.9%, and 97.2% recall, respectively, whereas the best models in the classical-only approach achieved 86.7%, 95%, and 89.7% intrusion detection accuracies, and 70.7%, 89.8, and 80.6% recall, respectively.</p>