基于LSTM-Attention的汽车CAN/CAN-FD协议入侵检测模型OA

In response to potential security risks in CAN/CAN-FD protocols,an LSTM-Attention intru-sion detection model was proposed.Input sequences were first processed by an LSTM layer to extract hidden states;these states were then weighted and summed through an Attention mechanism,after which a multi-layer perceptron performed the final classification.DoS,fuzzing,and spoofing attacks were evaluated on the CAN-FD Intrusion dataset,and comparative experiments with several models were conducted on the Car-Hacking dataset.The results show that the proposed model achieves preci-sion,recall,and F1-score above 99.9%for all three attack categories;it captures attack patterns more effectively,delivers superior detection performance,and exhibits strong generalization capability.