Research On RF Fingerprint Algorithm Based On Steady State Signal

The radio frequency fingerprint identification technology is a method for identifying different wireless devices through differences in the hardware of communication devices,and is intended to assist in solving the problem of secure access to wireless networks.In this paper,the fine fingerprints of the steady-state signal of the communication transmitter are extracted,and the fingerprint information is identified using classification technology.The main work of this article includes the following aspects:Firstly,introduced the basic model and principle of RF fingerprint identification,and analyzed it based on different hardware structures.On this basis,the recognition performance of different classifiers is introduced and three classifiers are selected for use in the study of this paper.Then,studied the RF fingerprint of steady-state signal based on pilotless structure.Extract the fingerprint features of pilot-free signals from different angles such as fractal features,higher-order moment features and phase noise spectrum.In fractal features,two features of box dimension and information dimension are studied,and the complexity of similar features is briefly described.Among the higher-order moment features,the three characteristics of R,J and estimated signal-to-noise ratio are studied,and the theoretical values of different higher-order moment features are briefly explained.In the phase noise spectrum characteristics,three different phase noise spectrum estimation methods are introduced,and the performance of different window functions is simulated.On this basis,combined with the measured data of the software radio platform,the extracted fingerprint features without pilot structure are classified and recognized in three classifiers.The test results show that the fingerprints extracted without pilot structure can identify the transmitter to different degrees.Among them,the fractal dimension and phase noise spectrum have better anti-noise ability,while the higher-order moments are greatly affected by noise.After this,studied the RF fingerprint of steady-state signal based on pilot structure.First,the commonly used carrier frequency estimation methods are introduced,and then the carrier frequency phase precise estimation method used in this paper is mainly explained based on the a priori pilot information.Several simple constellation features were extracted from the existing carrier frequency phase information,and the I/Q offset features of the constellation trajectory were studied.On this basis,each section combines the measured data of the test bench,and analyzes the recognition effects of different classifications when using the single feature of carrier frequency phase estimation feature,constellation feature and power spectrum statistical feature as the recognition input.The test results show that the signal fingerprints based on the pilot structure extracted in this paper can identify the transmitter to different degrees.Among them,the power spectrum statistical features have better anti-noise ability,but the overall recognition rate is lower.Although the carrier frequency phase estimation feature and the constellation feature are almost unrecognizable at low signal-to-noise ratio,the recognition effect is better when the channel conditions are better.Finally,build a RF fingerprint experimental test platform using the NI software radio platform USRP and its supporting software LabVIEW,and classified and verified the RF fingerprint features extracted in this paper.First,the software and hardware of the software radio platform are introduced in detail,and the communication module of the entire link is analyzed,including frame structure design,frame synchronization,and parameter configuration.Finally,the performance of RF fingerprint recognition is analyzed from three perspectives: single feature,multiple feature combinations and classifier performance.The hardware test results show that both the steady-state signal fingerprint without pilot structure and the fingerprint features extracted based on the pilot structure can achieve a good classification and recognition effect at high signal-to-noise ratio.Some features have strong anti-noise performance,including fractal features,phase noise spectrum,and power spectrum statistical features.Although some features are sensitive to signal-to-noise,they can achieve better recognition performance in a good channel environment.In the actual application process The two types of features can be used in combination.The performance of the Fine Gaussian SVM used in this paper is based on the extremely high computational cost and complexity in exchange for better recognition performance than the other two classifications.