Research On Physical Layer Authentication Technology In Wireless Communication System

With the rapid development of mobile communication technology,signal processing technology and Internet of Things technology,the role of wireless communication in existing communication systems continues to increase.However,due to the broadcast nature of wireless communications and the use of new network architectures,existing cryptography-based security schemes face the challenge of insufficient security strength,and physical layer security techniques,which start from the intrinsic security properties of the channel,can complement and enhance existing security mechanisms and are now an integral part of the security system for wireless communications.Among them,physical layer authentication technology,as a promising secondary authentication technology,aims to provide a low-complexity and high-security authentication scheme using physical layer features,and is one of the important components of physical layer security.This thesis focuses on the physical layer authentication technology in wireless communication systems,and its main contributions can be summarized as the following three aspects:Firstly,to address the problem of channel model and authentication model mismatch in the current millimeter-wave physical layer authentication technology,an eavesdropping channel model with spatial consistency for millimeter-wave frequency band modeling is designed based on the international common millimeter-wave channel model and the improvement of this model.Then,based on this channel model,a channel fingerprint-based attack detection scheme is proposed for identity spoofing attacks in 5G millimeter-wave MIMO systems.In the beam domain,the millimeterwave channel pattern exhibits beam sparsity and high directional characteristics,and this beam domain characteristic has extremely high correlation with the terminal location.This scheme uses the beam-domain channel pattern as a channel fingerprint and proposes an identity spoofing attack detection scheme based on this channel fingerprint;for the authentication problem therein,the terminal authentication problem in the spoofing attack is modeled as a binary classification problem for its channel fingerprint,and a supervised learning-based support vector machine algorithm is used to solve the classification problem.To obtain good classification results,similarity metrics such as Pearson correlation coefficient,cosine similarity,correlation matrix distance,and Euclidean distance are compared based on numerical analysis of channel fingerprints.Based on the comparison results,the optimal metrics are selected as classification features to train the classification model.The simulation results show that the scheme has high authentication accuracy and robustness even under low signal-to-noise ratio conditions.The attack detection accuracy is significantly improved compared with existing related mechanisms.Secondly,an authentication scheme based on device fingerprint and Convolution Neural Network(CNN)is proposed to address the problem that the relevant characteristics in the physical layer authentication technique based on device fingerprint are shifted due to temperature changes.When extracting the device fingerprint,a temperature sensor is used to obtain the device temperature in real time,and it is found that the instantaneous Carrier Frequency Offset(CFO)of the device is positively correlated with its temperature.Based on this characteristic,a temperature correction algorithm is designed to train the correspondence between temperature as the input of CNN network and the corresponding output as the instantaneous CFO,and to correct the CFO data based on the temperature correction algorithm.Finally,the authentication problem of device identity is modeled as a classification problem for fingerprints of terminal devices,and the problem is solved using a supervised learning-based CNN network.Experimental simulations using Universal Software Radio Peripheral(USRP)show that the authentication success rate of the CNN network is significantly improved after the data is calibrated by the correction algorithm.Finally,an application case in device fingerprint authentication is also presented to combine device fingerprint authentication technology with pseudo-base station detection in the field of pseudo-base station detection,and a pseudo-base station detection scheme based on device fingerprint is proposed to ensure the security of base station systems.Finally,a CNN-based physical layer authentication scheme is proposed for the authentication problem of multiple devices in Internet of Things(Io T)systems.Most of the devices in Io T systems are in stationary state,and their instantaneous Channel State Information(CSI)is strongly correlated with their locations,and the CSI characteristics of devices in different locations are very different.Therefore,a supervised learning method can be used to train the CSI data of known locations,so that the CNN model can learn the feature representations of different location nodes and associate them with the corresponding identity labels;when the training is completed,the CNN model can be used to authenticate the CSI data of unknown location devices,thus realizing secure access authentication and area intrusion detection of Io T devices.In addition,a detection mechanism based on the number of identity tags is proposed for the problem of near-access attacks.Since the number of legitimate devices in the Io T system is fixed,their corresponding identity tags are also fixed.If the number of identity tags detected by the CNN network exceeds the specified number,the presence of an attacker is determined.Simulation results show that the CNN network-based physical layer authentication scheme has good authentication performance.In particular,the authentication model can also be identified by a detection mechanism based on the number of identity labels in the face of near-access attacks.