Attacks to Automatous Vehicles: A Deep Learning Algorithm for Cybersecurity.

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Automotive Engineering

Background:

• Autonomous vehicles rely on Controller Area Network (CAN) bus for communication, creating vulnerabilities to cyber threats.
• The complexity of CAN bus data and traffic patterns facilitates unauthorized intrusions and various attacks.
• Rapid detection of message attacks is crucial for securing autonomous vehicle networks.

Purpose of the Study:

• To develop a high-performance artificial intelligence system for detecting cyber threats in autonomous vehicle networks.
• To enhance the security of the CAN bus protocol against intrusions and malicious attacks.
• To leverage deep learning for real-time identification of attack messages.

Main Methods:

• Utilized a real-world autonomous vehicle network dataset containing benign and attack (spoofing, flood, replaying) packets.
• Applied data preprocessing to convert categorical data into numerical format.
• Employed Convolutional Neural Network (CNN) and a hybrid CNN-LSTM model for attack message identification.

Main Results:

• The proposed deep learning models achieved high performance in detecting attack messages.
• The system demonstrated high accuracy (97.30%) in identifying various types of CAN bus attacks.
• The system showed superior performance compared to existing methods in real-time CAN bus security.

Conclusions:

• The AI-driven system effectively protects autonomous vehicle networks from cyber threats.
• Deep learning approaches, specifically CNN and CNN-LSTM, are highly effective for real-time CAN bus security.
• The proposed system offers enhanced detection and classification accuracy for improved vehicle network safety.