Machine Learning-Based Intrusion Detection System for Automotive Ethernet: Detecting Cyber-Attacks with a Low-Cost Platform.

Automotive Ethernet is being adopted in vehicles to provide the larger throughput that is required by autonomous vehicles. However, these vehicles may be subject to several cyber-attacks that compromise their operation and passengers' safety. This work proposes an Intrusion Detection System (IDS) that detects stream injection attacks on automotive Ethernet networks. The proposed IDS is based on feature generation and the XGBoost machine learning algorithm. Experimental results show that our proposed IDS achieves 0.9805 of AUCROC and a detection time of 620µs that allows real-time intrusion detection while using an inexpensive hardware platform, such as a Raspberry Pi. This is extremely important as cost is one of the automotive industry's main concerns.