Research On Single Antenna Intelligent Detection Algorithm For GNSS Spoofing

Satellite navigation signal has low emission power and is easy affected by unintentional and man-made interference during transmission.Among them,spoofing interference can make the receiver calculate wrong navigation and positioning result and even control the receiver,which is extremely harmful.In recent years,with the development of spoofing interference generation technology,its implementation is more flexible and cost-effective,and the case of successfully deceiving target receiver is increasing,broughting unsafe effects to many fields,so it is of great significance to detect spoofing interference.Single antenna spoofing intelligent detection algorithm based on machine learning was gived in this thesis.The main work of this thesis is as follows:Firstly,the detection method using a single parameter has certain limitations.Aiming at this problem,a multi-parameter detection method based on support vector machine and embedded in the receiver was gived.Spoofing process makes a series of parameters change.Carrier doppler frequency shift,pseudorange and other parameters obtained by the receiver were selected as the feature input of the C-support vector machine,constructing a classifier for detecting spoofing interference.C-support vector machine optimized by traditional grid search is liable to sink into the topical optimum,reducing the performance of the classifier.So cuckoo search algorithm was applied to optimize the C-support vector machine.Validated the effectiveness of the algorithm using both simulated and publicly available data.Experimental result shows that the method can effectively detect spoofing interference,and the C-support vector machine algorithm optimized by cuckoo search can further improve the accuracy and reduce the false alarm rate.Secondly,spoofing interference detection method using parameter needs to change the receiver,artificially selecting features after signal processing.A detection method based on neural network and externally connected to the receiver was gived subsequently.Cepstrum operation was performed on the received satellite signal,then it was used as the input of the one-dimensional convolutional neural network.Features were extracted autonomously by the network,and key features were enhanced using a multi-head self-attention mechanism,constructing a detection model from the signal end directly to the output end.Validated the effectiveness of the algorithm using both simulated and publicly available data.Experimental result shows that the method can effectively detect spoofing interference with high accuracy,and the improved network using the attention mechanism can further improve the accuracy.