Research And Implementation Of RF Fingerprint Identification Technology

Radio frequency(RF)fingerprinting is the unique RF signal characteristic of each RF device,similar to human fingerprints.It can be used as an ID for RF devices and has the advantage of being difficult to forge.It has broad application prospects in the fields of national defense security and civil communications.This thesis mainly studies the practical application and realization of RF fingerprint identification technology.The specific research results are as follows:(1)An RF fingerprinting method based on IQ-GRU is proposed.By constructing an IQ-GRU deep learning model,IQ-related features and timing-related features are extracted from IQ signals,and then the efficient and lightweight RF fingerprinting method is realized.The experimental test results show that,for 10 wireless routers,in the actual electromagnetic environment,the recognition accuracy rate reaches 96.65%,which is better than other similar methods,and the robustness is stronger.At the same time,compared with similar methods,this method has fewer parameters and shortens the training time by more than 70%.(2)An RF fingerprinting method based on channel equalization is proposed.By using the LMS adaptive filter combined with the time-domain training sequence of the signal frame header,the channel equalization and compensation of the signal to be identified are carried out.Then,the proposed IQ-GRU model is used to extract RF fingerprint features from the equalized signal for device identity recognition.Experimental test results show that this method can effectively mitigate the adverse effects of wireless fading channels on RF fingerprint recognition,and can improve the recognition accuracy rate from 62.1% to92.25% for signals affected by wireless fading channels.(3)An RF fingerprinting method based on domain adversarial network is proposed.By adding a domain discriminator to the IQ-GRU model,the signal samples received under different channels are mapped to the same feature space,which can improve the RF fingerprint recognition performance under wireless fading channels without knowing the signal prior knowledge.Experimental test results show that this method can achieve a recognition accuracy rate of 91.75% under wireless fading channels.(4)A low-power and low-cost RF fingerprint recognition system was designed and implemented.The embedded Zedboard development board was used as the main controller,combined with the RF chip AD9361 to achieve real-time reception and processing of RF signals.At the same time,the proposed IQ-GRU algorithm model was run on the Jetson Nano development board,and the GPU inside the board was used to accelerate the inference of the neural network,extract RF fingerprint features from the received signals and perform classification recognition.Experimental tests show that,compared with high-performance computers,the designed system reduces power consumption by 95%while ensuring running speed and recognition performance.