Research On Access Authentication Based On Radio Fingerprint Characteristics

In recent years,with the personalized development of mobile Internet,Internet of things and intelligent devices,and the increased demand for machine-to-machine(M2M)communication mode,massive terminal connection and data transmission require 5G fast and high security access becoming very challenging.Due to the high computational complexity of the traditional access authentication method based on unbreakable computation,the terminal cannot bear it.At the same time,the traditional security mechanism needs a longer key when the computing performance is gradually improved,and once the key is leaked,the consequences are unimaginable.Some authentication methods based on the essential characteristics of wireless devices become the enhancement solutions for the future network mass terminal and data access authentication.As the radio frequency fingerprint generated by the wireless transmitting device can only represent the wireless device,and has the physical characteristics that are difficult to clone,it can be used for identity authentication of wireless devices.The asymmetric authentication method may meet the requirements of high speed,low power consumption and high security of 5G network.The existing research on radiofrequency fingerprint authentication,the authentication rate is not high,which limits its application.Starting from improving its authentication rate,this paper studies the radio frequency fingerprint identification algorithm based on machine learning,radio frequency fingerprint transient signal feature extraction method,and radio frequency fingerprint feature dimensionality reduction method.there are five aspects of the work is completed as follows:1.The category separability measure is used as a new method to evaluate the performance of radio frequency fingerprint,and the feasibility of the method is verified by experimental simulation.2.A new feature extraction method for radio frequency fingerprint transient signals is proposed,which is called double maximum feature extraction algorithm.It is verified in Matlab that the performance of noise signal is better than wavelet transform.3.This paper discusses the method of radio frequency fingerprint identification based on machine learning,builds three models of radio frequency fingerprint identification based on machine learning algorithm,analyzes the influence of machine learning algorithm parameters on identification performance,and verifies the excellent performance of radio frequency fingerprint identification based on support vector machine.4.In order to solve the problem of high computational complexity and time consuming,a principal component analysis(PCA)algorithm based on multi-resolution analysis and ReliefF was proposed.Experiments show that the scheme can achieve better recognition performance with lower complexity.5.In view of the low signal-to-noise ratio(SNR)of radio frequency fingerprint identification,a feature transformation algorithm,symbol feature extraction is proposed.Based on this algorithm,a ReliefF and principal component dimensionality reduction algorithm for sample screening based on symbol feature is proposed,which can achieve high sample simplification and high recognition rate at low SNR.It is suitable for the terminal environment of lightweight recognition and authentication.