Research On Individual Identification Of Communication Device Based On Signal Fingerprint

With the further development of computer computing performance and the field of communication,the individual identification technology of communication device has a broader development prospect.This technology has application prospects in the field of civil security and electronic countermeasures in modern warfare.In recent years,many researchers have carried out extensive research on it.The individual identification of communication device mainly depends on the difference of signal fingerprint caused by the difference of hardware to distinguish different communication devices.Different communication devices have different signal fingerprints,so as to achieve the purpose of identification of the communication device.This thesis studies some technologies related to communication device identification,including feature extraction and classifier recognition.By extracting the different features of the signal as the signal fingerprint to identify the communication device and analyze the characteristics of different features.The communication devices are identified based on ensemble learning and deep learning,and the differences of different classifiers in the recognition rate and other parameters were compared.The main work includes the following aspects:(1)This thesis studies the feature extraction method for the combined integral bispectrum,The bispectrum of signal is extracted,and the visual analysis of bispectrum and its ability to suppress Gaussian white noise is carried out.The recognition rate of a single integral bispectrum and a combined integral bispectrum for communication device is compared,and the results show that the recognition rate of the combined integral bispectrum is more than 20%higher than that of the single integral bispectrum.Through GBDT and XGBoost classification model to verify the effect of combined integral bispectrum in mobile phone recognition.The results show that the recognition rate of the combined integral bispectrum in the process of mobile phone recognition can reach more than 96%.The robustness to amplitude interference is poor,the recognition rate is reduced by 7-8%under amplitude interference,and the recognition rate is reduced to 80%under multiple interferences.(2)This thesis studies the feature extraction method for wavelet packet energy spectrum.The wavelet packet energy spectrum of the signal is extracted after 7-level wavelet decomposition and analyzed visually.The GBDT and XGBoost classification models are used to verify the effect of identifying mobile phones.The results show that the recognition rate is about 83%when the energy spectrum of wavelet packet is used as the feature,and it is more sensitive to the classification of mobile phone models,less affected by the amplitude interference,and has better robustness,under the amplitude interference,the recognition rate only reduced by about 1%The overall recognition rate is lower than that of the combined integral bispectrum.(3)This thesis studies the communication device identification based on deep neural network.The new features are obtained by splicing the combined integral bispectrum and wavelet packet energy spectrum,and the new features are used to identify the communication device in the environment of deep learning.The results show that the recognition rate can reach 99.8%by using the new feature in deep neural network.It has good robustness under noise interference and amplitude interference,and the recognition rate is only reduced by less than 1%,which can effectively combine the advantages of the two features..At the same time,the recognition rate of deep neural network is significantly improved compared with ensemble learning,and the recognition rate of communication device reaches more than 98% under two kinds of interference.