Research On Radio Frequency Fingerprint Extraction And Identification Technology Of Mobile Terminal Devices Subtitle

Nowadays,wireless communication technology is developing rapidly,and the number of mobile terminal devices has exploded.The openness of mobile communication networks and the portability of mobile terminals have brought security risks to wireless communication systems.Traditional wireless network security system based on encryption algorithms and security protocols above the physical layer is increasingly revealing its drawbacks.Therefore,physical layer security technology is of great interest because it is a more secure,effective and cost-efficient method of device authentication.The radio frequency(RF)fingerprint feature based on the physical layer of the device is of great significance and practical value for securing wireless networks.RF fingerprint is based on the characteristics of the device’s own electronic components,which is unique,stable,and difficult to falsify and tamper with.Moreover,the identification of mobile terminal devices through the device’s RF fingerprint characteristics does not require modification of the system at the application level.Compared to other security methods,RF fingerprint is less expensive to implement and the complexity of the identification system is lower.Based on RF fingerprint technology,this thesis studies the uplink physical layer technology of both GSM and LTE systems,and designs RF fingerprint extraction and identification systems for mobile terminals based on GSM and LTE technologies respectively.And the effectiveness of the system for physical layer security systems is verified by experimental results.The main work and innovation points of this thesis are as follows:1)This thesis studies the uplink physical layer technology of GSM and LTE systems respectively,including the physical layer frame structure,random access process and baseband signal modulation method,and accordingly proposes the detection and capture algorithm for target frames of target devices.For GSM mobile phone devices,the transient starting point method of the GSM uplink burst signal and the synchronization and phase offset compensation technology for the burst signal are studied.The technique of extracting transient features of GSM signals based on discrete wavelet transform and GSM signal features based on Differential Constellation Trace Figures(DCTF)features are proposed.For Universal Software Radio Peripheral(USRP)devices as LTE terminal devices,the detection and capture of the preamble signal of the LTE uplink Physical Random Access Channel(PRACH)are studied,and the LTE signal feature extraction technique based on LTE preamble signal DCTF features is proposed.2)Based on the DCTF feature of the device,the DCTF feature generation mechanisms of the Gaussian Filtered Minimum Shift Keying(GMSK)modulation signal and the LTE PRACH preamble signal are theoretically deduced,and the effects of different nonlinear parameters on the DCTF feature distribution are investigated at the theoretical level.3)Based on Convolutional Neural Networks(CNN),the design of a DCTF-CNN based device RF fingerprint extraction and identification system is implemented,which combines the advantages of both DCTF and CNN without the need of the modulation method,signal synchronization or other priori information,by directly differencing the baseband signal and then drawing the DCTF,and then using the CNN model to train the classification,with low system complexity.4)Based on the DCTF-CNN RF fingerprint extraction and identification system,a complete set of hardware and software combined with the experimental procedure is designed to extract and verify the RF fingerprint features of GSM mobile terminal and LTE Universal Software Radio Peripheral(USRP)mobile terminal.For 6GSM mobile phone devices,the identification accuracy is higher than 85% when the signal-to-noise ratio(SNR)is higher than 15 dB,higher than 92.97% when the SNR is higher than 25 dB,and higher than 99.77% when the SNR is higher than 50 dB.For 8 USRP devices used as LTE user equipment(UE),the model identification accuracy is higher than 85% when the SNR is higher than 20 dB,and the model identification accuracy can reach 99.73% when the SNR is higher than 35 dB.