In-Vehicle Network Intrusion Detection System Based on Bi-LSTM

Abstract

In order to improve the accuracy of in-vehicle network anomaly detection by using the time correlation between CAN messages, this paper proposes an intrusion detection system based on Bi-directional Long Short-Term Memory (Bi-LSTM) network, and designed a sliding window strategy, Time series linear regression was performed on the real vehicle-mounted message data set to determine the sliding window of the optimal time length. The two-dimensional input data sample set is constructed according to the determined sliding window, and the two-dimensional data feature training classifier is learned by BILSTM network, and the trained network model is used to realize intrusion detection. In this experiment, four data sets were used to verify the detection performance. Compared with the existing research methods, the detection accuracy of the four data sets increased by 5.3%, 3.8%, 2% and 3.3% on average.