Deep Learning-Based Intrusion Detection System for Smart Cars

The increasing integration of smart and autonomous functionalities in modern vehicles has introduced new cybersecurity challenges, particularly concerning vehicular communication and control systems. This paper proposes a deep learning-based intrusion detection system designed to identify and classify cyberattacks in smart cars in real-time. The system leverages data collected from the vehicle’s CAN bus, sensors, and communication modules to detect anomalies associated with common attack types such as data spoofing, denial-of-service (DoS), signal jamming, and man-in-the-middle (MitM) intrusions. Three deep learning models—Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Long Short-Term Memory networks (LSTM)—are implemented and evaluated on a labeled dataset. The experimental results demonstrate that the LSTM model outperforms CNN and RNN in terms of accuracy (98.1%) and F1-score (97.9%), making it particularly effective for detecting sequential and time-dependent attacks. The proposed system exhibits strong adaptability, low detection latency, and potential for real-time deployment in connected vehicle environments, thereby contributing to the development of secure intelligent transportation systems.