A Machine Learning Approach Towards Secure In-Vehicle CAN Bus Network

In the automotive industry, different electronic circuit units (ECUs) communicate with each other using the controller area network (CAN) communication protocol. However, CAN bus attacks threaten the integrity and reliability of the transmission. In this paper, a software/hardware co-design using the PYNQ-Z1 FPGA board is suggested to protect data through the CAN interface. An architecture is implemented to detect attacks using a simple artificial neural network (ANN). ANN helps recognize whether there is an attack or not, thus enhancing the chances of detecting attacks on the CAN bus. Furthermore, a network address shuffling (NAS) security solution is implemented using the Verilog hardware description language. It raises opponent CAN data frame analysis costs. Experimental results show the effectiveness of our approach in distinguishing between legitimate traffic and malicious attacks.