Research On Physical Layer Security Technology Based On Neural Network

Due to the openness of the signal transmission in wireless communication systems,undesired receivers can also obtain the transmitted signals.The security of confidential information is a key issue in wireless communication systems.With the development of signal processing technology and channel coding technology,it is possible to realize secure transmission of information in the physical layer.Multi-antenna beamforming and artificial noise assisting technology are important technical means to achieve the physical layer security.This thesis studies the application of neural network in physical layer security signal processing.Neural network is used to realize adaptive security beamforming and the full-duplex legitimate receiver signal combining and self-interference cancellation.The main work of this thesis is as follows:1.The scheme of beamforming of signals and artificial noise using neural networks in the multi-input single-output system is studied.The channel estimation,channel state information feedback and beamforming design are combined into a reverse training process using linear neural network.Based on the reciprocity of the channel,the training sequence is sent by a legitimate receiver,and the secret signals and artificial noise beamforming weights of multi-antenna sender are obtained through neural network training.The bit error rate and secrecy rate are simulated and analyzed,the results show that the proposed scheme is feasible and effective,and the security performance of the scheme is very close to the performance of the conventional secure beamforming scheme under the deal channel state information.2.The scheme of realizing the self-interference cancellation and signal combining of the full-duplex legitimate receiver using neural networks in the single-input multi-output system is studied.The legitimate receiver receives the signal while transmitting artificial noise to interfere with the eavesdropper,but the artificial noise also enters the signal receiving channel of the legitimate receiver.Two neural networks are designed,one for the combination of the receiving antenna signals,and the other for reconstructing the self-interference according to the artificial noise transmitted by the transmitting channel,for self-interference cancellation of the received signal.The designed scheme is simulated and analyzed,the results show that the performance of the designed combiner is similar to that of the optimal combiner with accurate channel state information,and the effect of interference cancellation is also expected.