Vehicle Controller Area Network Inspection Using Recurrent Neural Networks

Abstract

The increasing connectivity in vehicles brings the potential for cyber-attacks, which can result in safety hazards or vehicle malfunctioning. Therefore, it is crucial to develop novel methods that can protect the vehicular network from malicious actors. In this paper, we propose a method of utilizing PCAP (Packet Capture) payloads extracted from the Controller Area Network (CAN) of a sample vehicle dataset, and applying a Recurrent Neural Network (RNN) to detect malicious or benign activity in the dataset. The process converts the extracted hexadecimal values of the payload to tensors and are labeled as either malicious or benign. The tensors are then passed through a neural network to produce the outputs. Our method showed to detect 78% of all packets within the dataset, which indicated its effectiveness in identifying cyberattacks in vehicular networks. This approach can eventually be applied to real-world scenarios, where the detection and prevention of malicious activity can make vehicles more secure.