Machine learning-based detection and mitigation of cyberattacks in adaptive cruise control systems.

Area of Science:

• Cybersecurity
• Automotive Engineering
• Machine Learning

Background:

• Vehicle-to-Vehicle (V2V) communication enhances Adaptive Cruise Control (ACC) functionality but introduces cybersecurity vulnerabilities.
• False Information Injection (FII) attacks can manipulate V2V messages, increasing collision risk and reducing driving efficiency.
• Existing ACC systems lack robust defenses against sophisticated V2V message manipulation.

Purpose of the Study:

• To investigate the impact of three distinct False Information Injection (FII) attack types on vehicle collision risk and driving efficiency.
• To develop and evaluate a novel machine learning-based onboard model, ACC anomaly Detection and Mitigation (ACCDM), for enhancing ACC system resilience.
• To validate the effectiveness of ACCDM in detecting and mitigating V2V-based cyberattacks in real-time.

Main Methods:

• Simulated three types of False Information Injection (FII) attacks on ACC systems.
• Developed a machine learning model (ACCDM) for real-time monitoring of vehicle parameters and anomaly detection.
• Conducted simulations under continuous and clustered attack scenarios to assess ACCDM's performance.

Main Results:

• ACCDM demonstrated high accuracy in detecting cybersecurity threats originating from V2V message manipulation.
• The proposed model successfully preserved safe following distances during simulated attacks.
• ACCDM effectively mitigated the negative impacts of cyberattacks on ACC system safety and driving efficiency.

Conclusions:

• The developed ACCDM model offers a robust solution for enhancing the cybersecurity of ACC systems against V2V communication threats.
• Machine learning-based anomaly detection is a viable approach for real-time threat identification and mitigation in connected vehicles.
• ACCDM contributes to safer and more efficient autonomous driving by strengthening ACC resilience against cyberattacks.