| Due to the openness of wireless channels,the access authentication of wireless communication devices is prone to threats such as identity impersonation attacks.However,traditional encryption authentication techniques may have high computational complexity and key leakage issues.How to effectively ensure the security of wireless communication device identity authentication has important research value.Radio frequency fingerprint features that can characterize wireless communication devices at the physical layer can be extracted by radio frequency fingerprint identification technology,which enables efficient and reliable identification of wireless communication devices,thereby completing secure authentication of wireless communication user identities.Therefore,The research on radio frequency fingerprint identification technology is of great significance,this article studies the extraction and recognition schemes of radio frequency fingerprints.In order to solve the problem of radio frequency fingerprint(RFF)that characterizes the essential characteristics of devices being submerged by noise.In this paper,an adaptive variational mode decomposition(AVMD)algorithm,is proposed for RFF extraction.The proposed algorithm uses the correlation coefficients between the decomposed modes and the proportion of the energy of each mode in the total energy of the signal to judge whether there is mode mixing or over-decomposition during the decomposition process.Based on that,the appropriate decomposition mode number and penalty factor are selected in an iterative manner.The experimental results show that compared with the original signal,the reconstructed signal obtained by AVMD algorithm not only retains relatively complete fingerprint features,but also has a certain degree of noise suppression.In order to solve the problem of limited fingerprint features extracted by a single network model,which leads to low identification accuracy.In this paper,a CNN-GRU parallel composite network model is proposed.This model combines the advantages of the convolution neural network(CNN)and Gate Recurrent Unit(GRU)to extract frequency and time domain features.Through the user-defined addition layer,time-frequency feature fusion is performed on the reconstructed signal obtained from AVMD algorithm to obtain more abundant RFF features,so as to better classify and identify wireless communication devices.By a user-defined addition layer,perform time-frequency domain feature fusion on the reconstructed signal processed by the AVMD algorithm to provide richer RFF information.This leads to improved classification and identification of wireless communication devices.this paper network model better than that of existing methods such as the CNN-LSTM series model,CNN,LSTM,and Bi-LSTM,with fewer network parameters,faster computational speed can achieve better results. |
