| With the development of Global Positioning System(GPS),it has been widely used in military and civilian fields,bringing great convenience to people.However,the long distance between the GPS satellite and the ground has caused the GPS signal to reach the near-Earth space user,the signal power has been greatly attenuated,and the signal is basically drowned in noise.Although the GPS signal under the influence of natural factors can still enable the receiver to achieve navigation and positioning functions,the openness of the civil GPS signal structure provides opportunities for lawless elements to spoof and jam with the GPS signal receiver,so it is important to detect spoofing jamming in the GPS system.In response to the endless spoofing jamming methods,this paper carries out the research of GPS spoofing jamming detection under the selfattentive mechanism,and the main work carried out is as follows:(1)The vulnerability of GPS signal is studied by analyzing the principle of GPS system,including GPS receiver,GPS signal structure and so on.A Matlab-based software receiver was developed to simulate the whole process of GPS signal capture and tracking.Based on this,the principles of two GPS spoofing jamming methods,forwarding type and generation type,are analyzed,and an example of primary generation type spoofing jamming using Hack RF is given.(2)By analyzing 1D Convolutional Neural Networks(1D-CNN),traditional Recurrent Neural Network(RNN)and RNN-derived network model Long Short-Term Memory(LSTM)structure,considering that GPS signals have time series correlation,these three types of models are applied to spoofing jamming detection.Based on this,we propose a CNN+LSTM model under the self-attentive mechanism,using 1D-CNN for local feature extraction and LSTM for global feature extraction to complete spoofing jamming detection by feature fusion and network fusion.(3)Based on the GPS spoofing jamming detection performance metrics,the Ratio detection metrics value for a continuous period of time is selected as the input of the neural network to complete the GPS spoofing jamming detection.The relevant data acquisition is realized by GPS software receiver,and the corresponding data set is generated and simulated and verified under four GPS spoofing jamming detection models,namely 1D-CNN,RNN,LSTM and the proposed CNN+LSTM under the self-attentive mechanism.The experimental simulation results show that the CNN+LSTM model under the self-attentive mechanism performs better in terms of convergence speed and detection success rate. |
