Research On 5G-oriented Physical Layer Authentication Technology

In order to meet the needs of vertical industries,the fifth generation of mobile communication(5G)standards include in many new communication scenarios and technologies,which however create many new security issues.5G security has become a research focus in recent years.Authentication,as one of the most important parts of communication security,has also drawn increasing attentions in recent years,due to its incredible lightweight characteristics.This thesis studies the 5G-oriented physical layer authentication technologies according to the new technologies,scenarios,and methods in 5G.The main contributions are given as follows:1.Research on Full Duplex Equipment Authentication Technology Based on Self-Interference ChannelFor the authentication of a full-duplex equipment,the characteristics of the self-interference channel are introduced into the physical layer authentication framework.Based on the maximum likelihood estimation and statistical analysis of the wireless signals,two different authentication schemes,respectively based on the self-interference channel estimation and the statistical value of the received signal are proposed.The former can provide fast authentication with relatively high energy cost,while the latter has lightweight performance,but requires a relatively long sample-collecting time.Based on the probability theory,a closed-form expression of the detection probability is derived,and the validity of the proposed schemes are verified by simulations.It is found that even if the attacker and the relay are at the same location,the detection probability can still reach more than 90%.2.Research on UAV authentication based on Doppler frequency shiftUtilizing the characteristics of the Doppler frequency shift generated by UAV mobile communication,the authentication of UAV communication is transformed into a hypothesis test problem,and two authentication schemes are proposed based on the real-time Doppler shift estimation and the period statistics of the Doppler shift respectively.The former can guarantee instant authentication with relatively poor performance,while the latter has higher accuracy,but only gives periodic monitoring results.The effects of system parameters,such as communication distance and movement direction,on authentication performance are simulated.It is found when the attacker is eight meters away from a legitimate transmitter,the detection probability can exceed 90%.3.Research on IoT authentication based on machine learningFor the numerous devices and difficult authentication in the Internet of Things scenario,a physical layer authentication technology based on machine learning is proposed.The supervised learning algorithm is used to integrate and train the channel state information of the IoT devices,and two physical layer authentication schemes based on the k-nearest neighbor(k-NN)algorithm and the logistic regression algorithm are proposed.Simulation results show that,through group authentication of IoT devices,the authentication efficiency of the physical layer authentication based on machine learning is three times higher than that based on hypothesis testing.