Research On Specific Emitter Identification Based On Radio Frequency Fingerprinting

The widespread application of wireless communication technology has brought profound changes to people’s live and work.However,compared with traditional wired networks,wireless networks are more vulnerable to malicious attacks from illegal users because of their openness.As a result,it is crucial to focus on their security issues.Specific emitter identification technology identifies and authenticates different emission sources by extracting subtle differences in the radio frequency signals of wireless radiation sources.This discrepancy is referred to as the "Radio Frequency Fingerprint" of the device and it is difficult to tamper or counterfeit.Therefore,Specific emitter identification technology is considered as a new method for the safety certification in wireless communication systems.The research on specific emitter identification technology has great theoretical significance and practical value.This paper focuses on the research of specific emitter identification based on radio frequency fingerprints.The main works are as follows:As the number of wireless devices increases in the network and manufacturing techniques advance,the differences in fingerprint features between devices become more subtle.In addition,in complex environments,radio frequency fingerprints of emitters are susceptible to noise and interference,further complicating of specific emitter identification.It has become challenging to obtain more accurate RF fingerprints that can describe the device’s fundamental characteristics.To address the problem,a specific emitter identification method based on improved diagonal integral bispectrum is proposed.First,an estimated bispectrum of the emitting signal is obtained,and the autocorrelation of each secondary diagonal parallel line is calculated to enhance the subtle features of the signal.Second,an improved diagonal-correlation locally-integral bispectrum(DCLIB)is proposed by adaptively determining the integration interval of the spectrum through the signal strength.Based on this,the DCLIB signal is further used as the radio fingerprint and a deep residual network is designed.for communication emitter identification.According to experiments on a public dataset,the proposed method outperforms existing methods and achieves good anti-noise performance.In actual scenarios,there are often a large number of signals from unknown radiation sources that make it impossible to obtain their fingerprint features in advance.The identification accuracy of radiation sources will be impacted if it still done using closed set recognition techniques.In this situation,it is essential to correctly identify known class radiation sources and distinguish unknown class radiation sources,completing open set identification of radiation sources.To address this the problem,an open set specific emitter identification method based on an improved residual network and fast clustering algorithm is proposed.First,a residual network is designed to extract the RF fingerprint features of the emitters.The features not only have lower dimensions,but also can make the feature distribution of the same device closer and different devices more dispersed,providing a solid framework for open set identification.Second,the impact of abnormal circumstances on classification accuracy is reduced by using a quick clustering algorithm to eliminate abnormal samples and misclassified samples that deviate from the distribution range of this class,and the clustering centers of various radiation sources are identified.In the identification process,the distance between the detected sample and the known cluster center is measured,and a judgment threshold is set to assess unknown radiation sources while classifying known radiation sources.The experimental findings demonstrate that the proposed approach has good applicability and practicality,and that its identification accuracy can be maintained at over 96%under various levels of openness.