Research On Physical Layer Secure Antenna Selection Technique In Millimeter-Wave MIMO Communications

Millimeter-wave is the key technology to drive the future rapid development of 5th Generation(5G)wireless communications.Due to the short wavelength of millimeter-wave,it is affected by the path loss.By making full use of the array gain brought by the ultra-large-scale antenna array,it can compensate for the path loss.And the small size of the antenna in the millimeter-wave system can accommodate large-scale antennas,so millimeter-wave combined with massive multi-input multi-output(massive MIMO)is the main development direction of future 5G communication.However,large-scale antennas are kept open for signal transmission at the same time will generate extremely high energy consumption.And the high-frequency devices and the large number of RF links required in millimeter-wave communication are costly.Antenna selection technology meets the demand for multiple antennas with fewer RF chains and adaptively controls the antenna switching,which can greatly reduce the unnecessary RF link cost and energy consumption.In addition,users not only put forward higher requirements for communication rates but also pay more attention to communication security under the new technology.The physical layer security technology takes the endogenous security properties of the wireless channel as the starting point to discover better security performance based on the physical characteristics of the wireless channel,which is a powerful complement to the upper layer security technology.Among them,the physical layer security antenna selection technology uses the state characteristics of the wireless channel,and can effectively improve the signal transmission security by selecting the antenna that maximizes the channel traversal secrecy capacity for the signal transceiver.To better utilize the millimeter-wave characteristics,the achievable secrecy performance of the physical layer secure antenna selection scheme under the millimeter-wave system needs to be further studied and evaluated.This thesis focuses on the problem of imperfect Channel State Information(CSI)for signal processing due to errors in channel estimation and delay in transmission processing in millimeter-wave multi-input multi-output(MIMO)eavesdropping channels,and a secure transmission framework with joint channel prediction and antenna selection is proposed,further more,the impact of imperfect CSI on the secrecy performance of millimeter-wave communication systems is analyzed.In order to improve the secrecy performance of millimeter-wave MIMO systems.The main work and contributions of this thesis are as follows:First,this thesis proposes a physical layer secure transmission framework with joint channel prediction and optimal antenna selection,and gives the specific implementation process of the joint secure transmission framework in Time Division Duplex(TDD)and Frequency Division Duplex(FDD)systems,respectively.In this joint framework,to characterize the time-domain correlation of time-varying channels,the autocorrelation function of the wireless channel in the time domain is derived,based on which the upper bound of the ergodic secrecy capacity achievable by the channel is further derived as a performance index.Second,to address the problem of estimation error and transmission processing delay in CSI used for secure antenna selection scheme,this thesis proposes a channel state prediction scheme using the temporal memory property of the Long-Short Time Memory(LSTM)network to mitigate the adverse effects of CSI imperfection on the results of antenna selection scheme.Simulation results show that the LSTM-based CSI time series prediction scheme has at least 50% improvement in root-mean-square error performance compared with the traditional nonlinear autoregressive network-based channel prediction scheme,which effectively reduces the prediction error.Therefore,the channel prediction scheme in this thesis can effectively mitigate the negative impact of channel estimation error and transmission processing delay on the system performance,thus obtaining near-perfect prediction CSI.Again,after obtaining the predicted channel state,this thesis proposes an optimal TAS scheme based on Support Vector Machines(SVM)with high classification accuracy for the problem of high computational complexity of the traditional optimization-based Transmit Antenna Selection(TAS)scheme.In addition,to address the problem of high overhead in training the classification model,this thesis improves the parameter update strategy of the SVM-based TAS scheme using Stochastic Gradient Descent(SGD)method to improve its training efficiency.The simulation results show that,with the experimental parameter settings in this thesis,the time used for training decisions in the SGD-based TAS scheme is about 10 s,and the selection accuracy is as high as 98%;while the classical SVM-based TAS scheme takes about 1200 s,and the accuracy is about 92%.It can be seen that using the SGD method can effectively improve the high training cost problem of the SVM-based TAS scheme without loss of confidentiality performance,which verifies the effectiveness of the proposed scheme in achieving confidentiality performance and improving training efficiency.Finally,since Gradient Boosting Decision Tree(GBDT)shows better performance advantages in practical system applications,while SGD brings greater efficiency improvement,but its achievable confidentiality performance is not optimal;to address this problem,this thesis proposes a GBDT-based TAS algorithm,which can achieve a compromise between confidentiality and efficiency,which can achieve near-optimal confidentiality performance on the one hand and a certain efficiency improvement compared with the traditional SVM algorithm on the other hand.The simulation results show that the GBDT-based TAS scheme takes about 400 s to train the decision,while the SVM-based TAS scheme takes about 1200 s,which verifies the effectiveness of the proposed scheme.Further,the proposed security scheme is extended to the massive MIMO system,and it is analyzed and verified that increasing the number of transmitting antennas has a gain effect on the system secrecy performance,and increasing the number of users has a negative effect on the system secrecy performance.