Research On Physical Layer Authentication Machine Algorithm Based On 5G Massive MIMO

Driven by the increasing and increasingly complex demand for information transmission,a new generation of mobile communication system technology emerges at the historic moment,namely 5G.Multiple-Input Multiple-Output(MIMO)technology can meet the requirements of various indicators.For those who use MIMO technology for information transmission,such as a relatively abundant degree of freedom in space,they can cooperate according to the corresponding downlink signals of more users,so as to interfere with the eavesdrovers' signals.Based on this,the rich physical layer information brought by Massive MIMO is used to conduct identity authentication.The specific research contents are as follows:(1)For the Massive MIMO communication model,a multi-cell scenario is utilized where there is an associated Racian fading between the base station and the user device.In order to improve the accuracy of channel parameters acquisition,a channel estimation method was proposed,in which the channel model consisted of a deterministic Line of Sight(LOS)path and a random NLOS component describing the real spatially correlated multipath environment.The asymptotic spectral efficiency of different channel estimators is analyzed.The numerical results show that the Minimum Mean Squared Error(MMSE)estimator is better than the element-wise MMSE(EW-MMSE)estimator and the Least Square(LS)estimator,and the results increase with the number of antennas.(2)Based on Massive MIMO communication scenario,a hypothesis testing method is proposed to validate the user's validity.Combining arrival time,receiving antenna Angle and large-scale fading factor as feature vectors establish the best estimator based on support vector machine(SVM)of Massive MIMO authentication model,by extracting the physical channel characteristic information for authentication "similarity".The simulation results show that the grid search method has a good effect on the authentication of medium and low speed users under the premise of optimizing the parameters of support vector machine.In the case of the presence or absence of the estimator component,the accuracy of the evaluation is greatly affected.With the increase of signal-to-noise ratio,the accuracy of weight verification also increases.(3)Aiming at the problem that the kernel function parameters of SVM model are difficult to select when the sample set is different,and the convergence rate of SVM algorithm is slow when the sample data set is large,a novel evolutionary algorithm,tabu search algorithm,is proposed,which has high computational accuracy,fast convergence rate and is easy to implement.Tabu search is more suitable for the process of parameter selection and avoids the endless loop of local evaluation.However,the process of using tabu search algorithm to optimize the model depends on the initial solution,so the quality of the initial solution will greatly affect the efficiency of weight evaluation.C-W algorithm based on tabu search is proposed to optimize the quality of the initial solution.The simulation results show that the C-W algorithm based on tabu search is better than the tabu search algorithm which generates random initial solution and the traditional SVM based on kernel function.