Icarecs-2026 - Smart Can Bus Security System Using Hybrid Two-Step Deep Learning Model

The Controller Area Network (CAN) bus plays a vital role in modern automotive and industrial communication systems by enabling efficient data exchange among multiple Electronic Control Units (ECUs). However, traditional CAN architectures lack built-in security mechanisms, making them vulnerable to cyberattacks such as spoofing, replay, fuzzing, and denial-of-service (DoS) attacks. Existing Intrusion Detection Systems (IDS) primarily focus on detecting malicious traffic but suffer from high computational complexity, delayed response time, dependency on cloud infrastructure, and limited real-time prevention capabilities. To overcome these limitations, a lightweight and intelligent CAN Bus Intrusion Detection and Prevention System (IDPS) is proposed, integrating AI-driven encryption and decryption mechanisms with a hybrid CNN-LSTM deep learning model. The system ensures secure communication between dual Arduino-based CAN nodes by detecting and actively blocking malicious intrusions generated by unauthorized nodes. It operates with low latency, reduced hardware overhead, and high detection accuracy, providing enhanced data confidentiality, integrity, and network resilience. This approach demonstrates significant potential for real-time applications in autonomous vehicles, industrial automation, and critical IoT networks, offering a scalable and adaptive solution for next-generation CAN Bus security.