Research On Wi-Fi Radio Frequency Fingerprint Extraction And Identification Technology

In recent years,with the popularization of mobile terminals in wireless communication networks,the identification and authentication technology of wireless devices has become a research hotspot.At present,the commonly used authentication schemes based on MAC address and digital certificate have a certain degree of security flaws.Faced with the security threats of traditional authentication mechanisms mentioned above,the research on the radio frequency fingerprint(RFF)extraction and identification technology in the field of physical layer security is conducted in this thesis.The radio frequency fingerprint can be utilized for device authentication to obtain higher security because RFF is a hardware characteristic formed by the differences between the electronic components of different devices,and it is difficult to forge.Compared with other radio frequency technologies,Wi-Fi systems are more widely used.However,there are still some deficiencies in current RFF extraction methods for Wi-Fi devices.Wi-Fi signals have wider spectrum bandwidth so they are more susceptible to channel multipath interference,and yet the features extracted in the existing methods are not highly distinguishable,resulting in a low recognition accuracy.In addition,the number of experiment samples and experimental scenarios in the current research are not rich enough,and there is insufficient research on the robustness of the radio frequency fingerprint during the long term operation of the devices.In view of the limitations of current research,the further research on Wi-Fi radio frequency fingerprint extraction and identification technology is conducted in this thesis.The Wi-Fi signals that conform to the orthogonal frequency division multiplexing(OFDM)physical layer specification in the IEEE 802.11 standard are the main research object in this thesis.Aiming to resist channel multipath interference and additive noise and extract stable and highly separable fingerprints of the devices themselves,some new Wi-Fi radio frequency fingerprint based character extraction and target classification and recognition methods are proposed.Subsequently,63Wi-Fi devices including multiple groups of the same brand and same model were used for classification testing in a variety of experimental scenarios.Finally,a security access control system for Wi-Fi devices was designed and built for practical testing,which verified that the methods proposed in this thesis have important application values in the field of wireless devices access authentication and identity recognition.The main work in this thesis is shown as follows:(1)The IEEE 802.11 protocol standard relevant to OFDM physical layer specification is studied,and the PHY preamble signal that can be used to extract stable fingerprint characteristics in the PPDU frame is analyzed.In order to solve the problem that channel multipath causes deviations in signal synchronization process and causes large fluctuations of the extracted RFFs,a more accurate time synchronization method is proposed and the other signal preprocessing procedures such as frequency offset estimation and phase offset estimation are improved to make the subsequently extracted RFF features unique and robust.(2)Aiming at the problem of existing channel multipath components and low discrimination in the Wi-Fi radio frequency fingerprint characteristics extracted by the existing methods,this thesis presents two Wi-Fi RFF extraction methods based on the combination of transient and steady state characteristics and power spectrum characteristics.The extracted features have abundant information and can better eliminate channel multipath interference and additive noise interference.Compared with the methods in the existing references,the methods mentioned in this thesis not only eliminate the channel characteristics,but also retain the stable and highly distinguishable fingerprint features.(3)Aiming at the radio frequency fingerprint features extracted by the above methods,the radio frequency fingerprint classification and recognition mechanisms based on statistical learning and sequential detection are proposed.Among the statistical learning classification algorithms,the Knearest neighbor algorithm which uses Euclidean distance as the measurement mode and the random forest algorithm show better results.Compared with the mechanism without sequential detection,using sequential detection recognition method can achieve a higher recognition rate under the condition of long term operation of the devices by tracking and observing the variation tendency of fingerprint characteristics over a period of time in the process of training and identification.(4)After the preceding comparative analysis,the better radio frequency fingerprint classification method is selected to carry out the experiment.Aiming at the problem that the number of experiment samples is small and the experimental scenes are not sufficient enough in the current research,a total of 63 Wi-Fi devices of five models are used in this experiment,and the number of devices of the same model is up to 20.In the static positions and moving states,the classification accuracy rates of these63 devices are 90.2% and 84.3%,respectively.In the experiment,the devices have been turned on at different times and worked continuously for a long time,and the classification accuracy shows no significant decline.(5)A set of Wi-Fi wireless devices security access system based on radio frequency fingerprint identification is designed and implemented,and experimental tests have been carried out in practical application scenarios.The system combines a variety of RFF features with the recognition mechanism of sequential detection,and finally realizes the admission control of the terminal devices.During the tens of thousands of tests on 20 experiment devices,the admission rate of legal devices and the interception rate of illegal devices reached 98.9%,which reflects the excellent recognition performance of the system and proves its strong practicability.