Security Access Authentication Technology Based On Physical Layer Channel Information

With the development of wireless communication technology and the increase of terminal devices,more and more information is transmitted in free space,the openness of wireless networks has brought great security risks to information transmission,that is,it is easy to be intercepted and attacked,the information security of wireless communication networks is increasingly prominent.Due to the development of quantum computing technology and electronic devices,the traditional cryptography-based information security technology has shown many deficiencies,many channel characteristics of physical layer have the advantages such as randomness and uniqueness,which have become one of the main research directions of physical layer security technology.Deep learning algorithms can study and analyze massive data,we can also easily obtain channel state information(CSI)by using channel estimation technology,therefore,our paper introduces deep learning algorithms to perform feature mining and regular analysis on a large amount of CSI data,so as to achieve secure access for users.The main contents of our paper are devided into two parts:1.Introduce deep learning algorithms and design physical layer security access authentication model and algorithms based on CSI: This paper establishes a security access authentication model composed of legal user Alice,illegal user Eve and receiving user Bob,and analyzes the feasibility and robustness of deep learning algorithm combined with the natural advantages of CSI.According to the characteristics and advantages of different deep neural networks,we introduce three kinds of deep neural networks: Convolutional Neural Network(CNN),skip-layer Convolutional Neural Network(skip-layer CNN)and Recurrent Neural Network(RNN).The convolution layer and poling layer of CNN make CNN have strong feature representation ability;skip-layer CNN reduces the computation and enhances the utilization of the feature map by reusing the feature map;RNN has a memory function that can analyze the contextual relationship of the data.For the entire system,the hidden layer of the deep neural networks can be regarded as the “blind extraction” process of CSI,and the output layer can be regarded as a classifier,which realizes the integrated design of authentication and classification.According to the comparison of simulation results,the conclusion that deep learning algorithms have better authentication performance than existing algorithms such as k-Nearest Neighbor(kNN)and Support Vector Machine(SVM)can be confirmed.In deep learning algorithms,skip-layer CNN has the highest authentication accuracy,RNN has the lowest time complexity.2.The algorithm verification platform is built based on USRPs: three USRPs are used as a legal user,an illegal user and a receiving user,one experimental system is built based on OFDM modulation technology.At the receiver,the channel estimation is performed and the measured channel data are collected,which are used to verify the proposed algorithms and some existing algorithms.The results show that the deep learning algorithms proposed in this paper have better authentication performance,which proves the feasibility of the physical layer security access authentication scheme in the actual communication system.