Auto-Updating Intrusion Detection System for Vehicular Network: A Deep Learning Approach Based on Cloud-Edge-Vehicle Collaboration

Intrusion detection systems play a crucial role in ensuring the safety of vehicle driving. Traditional intrusion detection systems face challenges in efficiently extracting key features from network traffic, resulting in inefficiency. To obtain better detection performance, many researchers adopted deep learning to build intrusion detection systems for vehicular network. However, these systems can only identify learned network attacks, while failing to identify the unknown attacks present in open scenarios. In this paper, we propose a deep learning-based auto-updating intrusion detection system for vehicular network, which uses a two-layer filter to effectively sense unknown attacks in open scenarios and proactively updates the attack recognizer to enable it with the ability to identify unknown attacks. In addition, we design a cloud-edge-vehicle collaborative model updating approach, where the updating task of the attack recognizer is deployed on a cloud server to reduce the computational load of the vehicle. The comprehensive experiments on three open scenarios show that the updated attack recognizer can accurately identify both known and unknown attacks.