利用CNN-LSTM融合模型实现GNSS诱导式欺骗干扰检测OA

Satellite navigation receivers have limited capabilities in countering induced spoofing attacks,and the traditional detection methods used are faced with challenges such as real-time processing difficulties and poor adaptability to preset discrimination thresholds.In view of this,the paper proposes a fusion neural network detection method based on CNN-LSTM.Firstly,the correlation peak aliasing characteristics during the spoofing pull-off phase was analyzed.Then,the ResNet-18 was used as the backbone of the convolutional neural network(CNN)to extract spatial features in the code phase domain and Doppler domain,and the long short-term memory(LSTM)network was employed to track the temporal dependencies across consecutive frames,so as to detect the inducing behavior of deceptive signals.To simulate the induced spoofing process,a correlation ambiguity function(CAF)sequence dataset was constructed to verify the detection performance of the fusion model.Experiments show that the detection accuracy rate of the proposed method for induced spoofing attacks exceeds 98%,which is improved by 2%than that of the traditional single models.Moreover,both the detection duration and model complexity can meet the requirements of civilian receivers.To sum up,it is an effective method in the field of anti-spoofing application of satellite navigation.