| |
| With the rapid development of intelligent driving technology and in-vehicle infotainment systems, the number of communication interfaces to the outside world has increased, and the architecture of the in-vehicle network has become highly complex. As a result, the Controller Area Network (CAN), which lacks security features, is more vulnerable to external attacks. Deep learning models are widely used in intrusion detection techniques, but it requires a large amount of data. However, accurate CAN attack data is not readily available in large quantities. To solve this problem, we propose an intrusion detection system (IDS) for in-vehicle CAN bus using deep transfer learning and support vector machines (SVM). In our IDS, the deep transfer learning model extracts the features of CAN messages; the SVM uses these features to identify if the CAN bus has been hacked. Experimental results show that with a limited amount of datasets, the proposed IDS still has a good detection performance with an accuracy of up to 99.1\%. |
|
| Keywords:Computer Application Technology; In-Vehicle Network Security; Controller Area Network; Intrusion Detection System |
|
