Intrusion Detection for CAN Using Deep Learning Techniques

With the advent of Internet of Vehicles (IoV), cars and commercial vehicles represent a convenient attack surface for cyber attacks. Many automobiles use the Controller Area Network (CAN) bus for internal communication. CAN is known to be susceptible to various types of cyber attacks. One constraint on intrusion detection systems (IDS) for CAN is that they need to be efficient due to lack of resources and the high traffic on a typical CAN network. This paper presents an implementation of simple 1D Convolutional Neural Network (CNN), Long Short Term (LSTM) and Gated Recurrent Units (GRU) networks on a recent attack data set for CAN. All models thus developed outperformed the existing state-of-art and achieve an almost perfect F1-Score of 1.0.