Research On Anomaly Detection Method Of Electromagnetic Environment Based On Deep Learning

With the rapid development of wireless communication technology,the use of a large number wireless communication equipment makes the scarce wireless spectrum resources more and more crowded.The electromagnetic environment anomaly detection method can provide technical support for monitoring and standardizing the compliant use of spectrum resources by a large number of frequency equipment.In addition,network security has become a global problem.Electromagnetic environment security detection in wireless physical layer is an important part of network security defense.The electromagnetic signals generated by the existing communication methods are easy to be intercepted or penetrated.If the types of abnormal electromagnetic signals can be detected,the active defense ability of the communication system against the abnormal electromagnetic signals will be improved.Therefore,how to effectively detect and judge the types of electromagnetic environment anomalies from the complex electromagnetic environment is of great significance to the monitoring of electromagnetic environment and the improvement of network security.In this paper,the method of electromagnetic environment anomaly detection based on deep learning is improved and modeled.The main contents can be summarized as follows:Firstly,this paper analyzes the characteristics of electromagnetic environment anomaly and summarizes the difficulties of electromagnetic environment anomaly detection.At the same time,the definition and mathematical model of electromagnetic environment anomalies and their types are defined.Secondly,the basic theory of deep learning is introduced,focusing on one-dimensional convolutional neural network and automatic encoder.Then,the existing automatic encoder anomaly detection model focuses on the reconstruction of power spectral density data estimation and ignores the local dynamic characteristics of the electromagnetic spectrum,which leads to the accuracy of the detection method based on the automatic encoder to be improved.Therefore,this paper proposes an anomaly detection model based on convolutional automatic encoder.The simulation results show that compared with the traditional auto-encoder method,the proposed method improves the detection accuracy of weak abnormal electromagnetic signals with different intensities.When the signal to interference ratio is 28 d B,the area under the ROC curve of different kinds of electromagnetic environment detected by the convolutional auto-encoder model increases by more than 0.02.At last,aiming at the problem that the current detection methods of electromagnetic environment anomaly do not classify the types of anomaly,the types of electromagnetic environment anomaly are defined from the perspective of spectrum occupancy.In this paper,two kinds of classification models of electromagnetic environment anomaly based on one-dimensional convolution neural network are proposed.The simulation results show that the algorithm based on one-dimensional convolution neural network can effectively classify the electromagnetic environment anomalies,which is more accurate than the previous deep learning anomaly sequence detection network.When the signal to interference ratio is 20 d B,the accuracy of anomaly classification reaches 95%.