Wireless Physical Layer Authentication Research Under Agricultural Environment

In recent years,smart agriculture and agricultural Internet of Things have gradually attracted people’s attention.For convenience,smart agriculture and agricultural Internet of Things use a large number of wireless communication devices.However,while the openness of these wireless communications brings convenience,it also brings great security risks to the transmission of information,which affects the security of agricultural communications.On resource-constrained agricultural equipment,information security technologies based on traditional cryptography have shown many shortcomings,such as the inability to support encryption algorithms with high computational complexity,difficulty in resisting brute force attacks,and easy disclosure of keys.In order to improve the security in the process of agricultural wireless communication,this paper uses PLA(Physical layer authentication,physical layer authentication)to prevent spoofing attacks,that is,CSI(Channel State Information,channel state information)is used for physical layer authentication.CSI is unique in space and time,and through the analysis and processing of CSI by machine learning and deep learning,legal and illegal transmitters can be well identified.In the existing research,the accuracy of physical layer authentication in a static environment is high,but the accuracy in a dynamic environment usually cannot meet the application requirements.Therefore,this paper conducts research on physical layer authentication based on deep learning for dynamic communication scenarios in an agricultural environment:(1)An improved CSI-VGG16 deep convolutional model is proposed to achieve physical layer authentication in the dynamic environment of agriculture.At the same time,a physical layer authentication algorithm based on Support Vector Machines(SVM),KNearest Neighbor(KNN),and Convolutional Neural Network(CNN)was introduced for comparative analysis;and compared with the traditional VGG11,VGG16,VGG19 models.(2)This paper proposes two CSI data enhancement algorithms based on CGAN(Conditional Generative Adversarial Network,Conditional Generative Adversarial Network)-FC-CGAN and RNN-CGAN.After learning and enhancing the CSI feature distribution in the dynamic agricultural environment,the generated data is mixed with the original data,and then CNN and CSI-VGG16 models are used to verify the accuracy of physical layer authentication.(3)Using the USRP(Universal Software Radio Peripheral,Universal Software Radio Peripheral)platform combined with Lab View software to build an OFDM signal transceiver system.In the agricultural environment,for static and dynamic plant factories and experimental fields,the channel data of legal senders and receivers and attackers in four different scenarios are collected,and the algorithms proposed in this paper are verified.