A Transfer Learning Based Intrusion Detection System for Internet of Vehicles

Abstract

With the fast expansion of the internet of vehicles (IoV) and the emergence of new types of threats, the traditional machine learning-based intrusion detection systems must be updated to meet the security requirements of the current environment. Recently, deep learning has shown exceptional performance in IoV intrusion detection. However, deep learning-based intrusion detection system (DL-IDS) models are more fixated and dependent on the training dataset. In addition, the behavior changes with the occurrence of attacks. They pose a real problem for the DL-IDS and make their detection more complicate. In this paper, we present a deep transfer learning based intrusion detection in-vehicle (TRLID) model for IoV using the CAN bus protocol. In our proposed model, a data preparation approach is proposed to clean up bus data and convert it to an image for usage as input to the deep learning model. Indeed, we used transfer learning characteristics because they enable us to transfer the source task's knowledge to the target task. Therefore, we trained our model using different dataset including different attacks. The experimental results show that our proposed TRLID achieved good results where the intelligence integration of transfer learning was efficient for attacks detection.