A Multiple Rényi Entropy Based Intrusion Detection System for Connected Vehicles

Area of Science:

• Cybersecurity
• Information Theory
• Network Security

Background:

• Vehicular networks face increasing cybersecurity threats.
• Existing intrusion detection systems (IDS) may not adequately address the complexities of vehicular communication protocols.
• Rényi entropy offers a generalized framework for information measurement, potentially enhancing anomaly detection.

Purpose of the Study:

• To develop and evaluate a novel intrusion detection system (IDS) for vehicular networks.
• To leverage the estimation of Rényi entropy with multiple orders for improved attack detection accuracy.
• To specifically target network intrusions within Controller Area Network (CAN) bus systems.

Main Methods:

• Utilized an efficient Rényi entropy estimation method for arbitrary real orders.
• Implemented a multi-order Rényi entropy estimation (orders 2, 3, and 4) based on Controller Area Network (CAN)-IDs.
• Employed a retrospective sliding window method for attack decision-making based on estimated entropy values.

Main Results:

• The proposed IDS demonstrated high accuracy in detecting intrusions.
• Achieved simultaneous false negative and false positive error rates below 1%.
• Validated performance using a recognized CAN-ID attack dataset.

Conclusions:

• The multi-order Rényi entropy-based IDS is a promising approach for securing vehicular networks.
• The method provides a robust and accurate solution for detecting network attacks in automotive systems.
• The low error rates indicate the system's reliability in real-world applications.