Nowadays, intelligent and connected vehicles (ICVs) have rapid development, and their cyber security problems have become increasingly significant. The cyber-attacks on ICVs directly threaten the property, privacy and even personal safety of drivers and passengers. Most existing intrusion detection methods are primarily designed for CAN bus, with relatively little research focused on automotive Ethernet. This paper takes SOME/IP as an example and focuses on attack detection in automotive Ethernet. An intrusion detection method combining time series prediction and rule-based approaches is proposed. An in-vehicle network traffic prediction model that combines CNN-BiLSTM with attention mechanisms is utilized to learn the temporal characteristics of data packets in automotive Ethernet. The cosine similarity between real messages and predicted results is utilized as a rule to identify attacks. Experimental results demonstrate the effectiveness of the proposed method.
An Intrusion Detection Method for Automotive Ethernet Networks Based on Time Series Prediction and Rule Setting
2024 IEEE Smart World Congress (SWC) ; 1820-1827
2024-12-02
1234411 byte
Conference paper
Electronic Resource
English
Reliability in automotive ethernet networks
IEEE | 2015
|TIBKAT | 2015
|TIBKAT | 2017
|TIBKAT | 2021
|