On the Application of Entropy Measures with Sliding Window for Intrusion Detection in Automotive In-Vehicle Networks

Area of Science:

• Cybersecurity
• Automotive Engineering
• Information Theory

Background:

• Modern vehicles face increasing cybersecurity risks due to enhanced connectivity and automation.
• Attacks on in-vehicle networks are feasible, necessitating robust intrusion detection systems.
• In-vehicle systems have unique constraints: limited computing power, data transfer limitations, and cryptographic system management challenges.

Purpose of the Study:

• To propose an effective intrusion detection approach for in-vehicle networks considering existing constraints.
• To evaluate the performance of various information entropy measures for detecting cyberattacks.
• To analyze the impact of hyperparameters on detection accuracy.

Main Methods:

• An information entropy-based method utilizing a sliding window approach for real-time detection.
• Evaluation of multiple entropy measures including Shannon, Renyi, Sample, Approximate, Permutation, Dispersion, and Fuzzy Entropy.
• Testing on a large public dataset of Controller Area Network (CAN) bus traffic with simulated Denial of Service, Fuzzy, and spoofing attacks.

Main Results:

• The sliding window entropy-based method demonstrates high detection accuracy for in-vehicle network attacks.
• The approach is time-efficient and does not require complex cryptographic systems.
• Specific entropy measures and hyperparameter choices significantly impact detection performance.

Conclusions:

• The proposed entropy-based sliding window method offers a viable solution for automotive cybersecurity.
• It effectively addresses the unique constraints of in-vehicle networks.
• Further research into optimal hyperparameter tuning can enhance attack detection capabilities.