ECU s Fingerprint Identification Based on CAN Bus Using 1D-CNN

Traditional Electronic Control Units (ECUs) fingerprinting classification methods for Controller Area Network (CAN) systems rely on manually defined features such as maximum voltage and average voltage, requiring a large number of features to effectively distinguish between different ECUs. However, these methods often face challenges in scalability and generalization across diverse environments. This study addresses these issues by adopting a one-dimensional Convolutional Neural Network (1D-CNN) model with automatic feature extraction capabilities, significantly reducing the reliance on manual feature engineering. The model features a short training cycle and fast inference speed, making it highly suitable for real-time applications in resource-constrained environments. Experimental results demonstrate that the model achieves a classification accuracy of 99.98%. Moreover, this approach shows great potential for deployment on embedded platforms and adaptability to dynamic operating conditions, paving the way for future advancements in vehicle security systems.