Deep Transfer Learning Based Intrusion Detection System for Electric Vehicular Networks

Area of Science:

• Cybersecurity
• Network Security
• Automotive Systems

Background:

• Controller Area Network (CAN) bus is crucial for In-Vehicle Networks (IVN) but faces security vulnerabilities due to complex architectures.
• Increasing cyber threats necessitate advanced intrusion detection systems (IDS) beyond traditional machine learning models.

Purpose of the Study:

• To propose a novel deep transfer learning-based IDS for enhanced cybersecurity in IVNs.
• To improve the detection accuracy of cyberattacks in real-time vehicle networks.

Main Methods:

• Developed a deep transfer learning-based LeNet model for IDS.
• Implemented effective attribute selection for identifying malicious CAN messages.
• Evaluated the model using real-world IVN data.

Main Results:

• The proposed IDS demonstrated superior detection accuracy compared to mainstream machine learning, deep learning, and benchmark deep transfer learning models.
• Achieved improved performance in detecting normal and abnormal activities in IVN.

Conclusions:

• The deep transfer learning approach offers a robust solution for real-time IVN security.
• The proposed IDS effectively addresses the evolving security challenges in modern vehicles.