Research On Individual Identification Technology Of Wireless Devices Based On RF-DNA Fingerprinting

With the development of the communication technology,how to effectively obtain device information is of great significance whether in the wireless network security or military reconnaissance.Under the complex electromagnetic environment,people not only focus on the basic parameters of signal features,such as frequency,modulation mode and so on,but also attach more and more importance to the subtle characteristics produced by the differences of individual parameters and production debugging of sources.According to these subtle characteristics,we can judge which device the signal comes from,and obtain the user's identity,so as to supervise the target electronic communication devices.In this paper,fingerprint extraction methods for wireless devices based on Radio Frequency-Distinct Native Attribute(RF-DNA)are studied,the main research are carried out as follow:Firstly,the basic theory and methods of individual identification of wireless devices are studied.It includes the generation mechanism of Radio Frequency Fingerprints(RFFs),signal preprocessing,feature dimensionality reduction and the classification algorithm.This section focuses on start point detection algorithm of the transient signal,the simulation results of variance trajectory detection and fractal Bayesian change point detection algorithm are given.The validity of the two detection methods is verified by the detection deviation of the measured signal under different Signal-to-noise Ratio(SNR)conditions.Secondly,the fingerprint feature extraction method of wireless devices based on RF-DNA is studied.Aiming at the transient signal of wireless devices,this paper makes a thorough study on the instantaneous characteristics of signal in time domain,the characteristics of wavelet coefficients and the characteristics of time-frequency distribution.Statistical features are introduced to construct RF-DNA fingerprints based on the transient signal.For the steady-state signa,this paper mainly studies the high-order statistical features.Integral bispectrum algorithm is used to construct RF-DNA fingerprints based on steady-state signals.The simulation results show that the feature extraction algorithm based on time-domain and time-frequency distribution can effectively separate 10 walkie-talkie devices in the ideal white Gaussian noise channel environment.The identification rate of the rectangular integral bispectrum feature in the high-order spectrum feature is close to 90%.Finally,the fusion method based on D-S evidence theory is studied.Based on the research of fingerprint feature extraction methods mentioned above,aiming at the shortcomings of traditional device identification based on single signal fragment,this section extracts a transient steady-state multi-fragment joint identification model based on D-S evidence theory.The simulation results show that the fusion method proposed in this paper has greatly improved the identification rate of different devices.In addition,based on the proposed fusion identification model,this paper builds an individual identification demonstration system for wireless devices based on USRP2930,and verifies the performance of the demonstration system through system functional testing.