Technical Research About Physical Layer Authentication Based On Wireless Channel Timing Characteristics

Wireless communication environment security is the fundamental guarantee of individual privacy and national property security.Prior to the arrival of 5G networks,traditional physical layer authentication(PLA)schemes have exposed the shortcomings of security authentication,due to the surge of the network mounts and the improvement of computing capabilities gradually.Even more,with the complexity and dynamics of the network environment,existing authentication methods based on wireless channel characteristics cannot adapt to the time-varying characteristics effectively,thereby failing to guarantee the reliability of security authentication.In order to extract the characteristics of the wireless channel fully and achieve reliable and safe lightweight PLA,this paper focuses on the PLA scheme based on wireless channel time feature and machine learning methods.The main research works are as follows:1)Firstly,this paper studied whether the characteristics of wireless channels can meet the reliability of PLA under specific conditions deeply.Based on the analysis of the single channel feature,the extraction methods of multidimensional spatial channel feature are pointed out,it changes the shortcomings of limited thresholds and affection of channel estimation bias in general scenarios,and improves the robustness and reliability.In addition,in view of the disadvantages of multidimensional feature authentication schemes,such as uncertainty and unknown dynamics,which are limited to the modeling of the identity authentication process,the effect of time series features on the identity recognition rate is studied,and two models are improved based on the predecessors.2)Under hypothetical conditions that only the wireless channel characteristic samples between the legitimates are available,and the samples of the illegal attacker are unknown,the existing PLA schemes are mostly based on binary assumptions to model authentication problems,that is,the problem is considered as classification of the channel feature conceptually.The modeling method of this problem has limitations,such as model timeliness and training sample balance.So,it cannot achieve the security authentication of a single and small sample.Therefore,this paper proposes a long short-term memory network(LSTM)based on dynamic prediction PLA scheme.This scheme solves the shortcomings of samples small,single and long-term failure effectively.Realizing PLA by predicting and comparing the unknown channel characteristic collected in next moment.The validation of the measured data proves the superior performance of the scheme in the case of limited samples.3)On the basis of existing PLA schemes which based on machine learning algorithms,this paper combines support vector machine algorithm(SVM)and dynamic extraction of channel timing features to design schemes.Specifically,the PLA system model of the scheme is established firstly,and a brief overview of the authentication process is proposed.Then,a new method of channel timing feature extraction(The feature matrix dynamic reconstruction scheme)is introduced,and PLA scheme based on SVM was presented.Finally,the effectiveness of the scheme was verified by using measured data under different parameter settings.Compared with the existing schemes which based on SVM and convolutional neural network(CNN),the performance of security authentication we motioned has been greatly improved.