Research On Radio Frequency Fingerprint Extraction Based On Differential Constellation Figure

With the rapid development of communication technology,wireless networks have covered various aspects such as national defense,economy,society,and people's livelihood,etc,and have been inseparable from people's daily lives.Wireless network technology enables people to get rid of the shackles of space and time,greatly improves the quality of life of people,and promotes the development of human society.However,the wireless network's information security is also emerging,which makes the demand for wireless device identification increasingly prominent.The traditional security policies are mostly established at the application layer.For example,the current authentication mechanisms generally use passwords and security protocols.However,the cryptographic mechanism has the risk of key leakage,and the security protocols usually have defects.Radio frequency fingerprints(RFF)generated by wireless devices have physical characteristics that are hard to be cloned.Using RFF to identify different devices is a physical layer method for protecting communication system security.Just as different people have different fingerprints,different wireless devices also have different RFF,which can be used for identification and access authentication of wireless devices.This paper reviews the research progress of RFF extraction technology in the past decades and analyzes two kinds of existing RFF extraction schemes based on modulation errors.Then proposes a new RFF extraction scheme based on the Differential Constellation Figure.First,this paper expounds the mechanism of how RFF generates,and summarizes the characteristics of RFF.Then summarizes the typical process and research status of RFF extraction.Two kinds of existing RFF extraction schemes based on modulation errors are analyzed: the Passive RAdiometric Device Identification System and the RFF extraction scheme based on the Constellation Trace Figure.The PARADIS program requires extremely sophisticated instruments as a receiver,and this program selects many parameters as RFF so that its calculation is complicated.The RFF extraction scheme based on the Constellation Trace Figure uses oversampling so that the computation increasing,furthermore,this scheme introduces two variables.Finally,this paper proposes a new RFF extraction scheme based on the Differential Constellation Figure.We implement the differential processing on the received baseband signal to obtain a stable differential constellation.Then the k-means clustering algorithm is used to get the different clusters.The intra-class average distance of clusters is combined with the summation of the intra-class distance is the RFF.This program has four advantages: 1)the program uses a relatively inexpensive software radio as the receiver.2)the program reduces the receiver sampling rate so that reduce the amount of data calculation,3)the program does not introduce introduces variables which may have impact on the system performance,4)the program uses the intra-class distance combined with the intra-class average distance of clusters after clustering as RFF.The fingerprint does not change with the distribution of the cluster center points.Finally,this paper uses USRPs to collect signals,and verify the reliability and practicality of this scheme by experiment using the two classification algorithms of C4.5 decision tree and support vector machine.The result shows that this scheme can obtain the RFF of the device without prior information and can still maintain good classification accuracy under different SNRs.When the SNR is higher than 6d B,the average classification accuracy can reach more than 90%.When the SNR is higher than 30 d B,the average classification accuracy can reach more than 98%.