CN-121980359-A - Open set radio frequency fingerprint identification method and device, storage medium and electronic equipment

CN121980359ACN 121980359 ACN121980359 ACN 121980359ACN-121980359-A

Abstract

The invention provides an open set radio frequency fingerprint identification method, an open set radio frequency fingerprint identification device, a storage medium and electronic equipment, wherein the method comprises the steps of calculating the class center distance of a target radio frequency fingerprint feature under each monitored object class, calculating the tail probability of the target radio frequency fingerprint feature under each monitored object class based on the class center distance of the target radio frequency fingerprint feature under each monitored object class and the target Cauchy probability accumulation distribution under each monitored object class, determining the target class probability of the target radio frequency fingerprint feature under each monitored object class by adopting the tail probability of the target radio frequency fingerprint feature under each monitored object class, and determining the radio frequency fingerprint identification result of the target radio frequency fingerprint feature based on the target class probability of the target radio frequency fingerprint feature under each monitored object class, wherein one radio frequency fingerprint identification result support is used for indicating whether the existing monitored object class exists or not. The embodiment of the invention can improve the accuracy of the radio frequency fingerprint identification.

Inventors

ZHENG YUNFEI
WANG WEI
GONG JING
ZHENG GUOJIANG
ZHANG TIANYI
WANG HAO

Assignees

航天时代低空科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (10)

1. An open set radio frequency fingerprint identification method, comprising: Acquiring a target monitoring object identity signal, and extracting characteristics of the target monitoring object identity signal to obtain a target radio frequency fingerprint characteristic of the target monitoring object identity signal; Determining class centers of all monitoring object classes in a monitoring object class set, and calculating class center distances of the target radio frequency fingerprint features under all the monitoring object classes based on the class centers of all the monitoring object classes respectively; Calculating the tail probability of the target radio frequency fingerprint feature under each monitoring object category based on the category center distance of the target radio frequency fingerprint feature under each monitoring object category and the target cauchy probability cumulative distribution under each monitoring object category, wherein the tail probability of the target radio frequency fingerprint feature under one monitoring object category is calculated by substituting the category center distance of the target radio frequency fingerprint feature under the corresponding monitoring object category into the target cauchy probability cumulative distribution under the corresponding monitoring object category; and determining a radio frequency fingerprint identification result of the target radio frequency fingerprint feature based on the target class probability of the target radio frequency fingerprint feature under each monitoring object class, wherein one radio frequency fingerprint identification result support is used for indicating whether the target radio frequency fingerprint feature is the existing monitoring object class or not.
2. The method according to claim 1, wherein calculating the class center distance of the target rf fingerprint feature under the respective monitor object class based on the class centers of the respective monitor object classes, respectively, comprises: Traversing each monitoring object category in the monitoring object category set, and taking the currently traversed monitoring object category as the current monitoring object category; Based on the class center of the current monitoring object class and the target radio frequency fingerprint feature, respectively calculating the current class distance of the target radio frequency fingerprint feature in each of a plurality of distance calculation modes, wherein the current class distance of the target radio frequency fingerprint feature in one distance calculation mode is the distance between the target radio frequency fingerprint feature calculated according to the corresponding distance calculation mode and the class center of the current monitoring object class; Determining a target weight set under the current monitoring object category, and carrying out weighted summation on the current category distance of the target radio frequency fingerprint feature under each distance calculation mode according to the target weight set to obtain the category center distance of the target radio frequency fingerprint feature under the current monitoring object category; and after traversing each monitoring object category in the monitoring object category set, obtaining the category center distance of the target radio frequency fingerprint feature under each monitoring object category.
3. The method according to claim 1 or 2, wherein determining the target class probability of the target rf fingerprint under the respective monitor object class using the tail probability of the target rf fingerprint under the respective monitor object class, respectively, comprises: Determining correction class probabilities of the target radio frequency fingerprint features under the monitoring object classes by adopting tail probabilities of the target radio frequency fingerprint features under the monitoring object classes respectively; And calling a target normalization function, and converting the corrected category probability of the target radio frequency fingerprint feature under each monitoring object category into the target category probability of the target radio frequency fingerprint feature under each monitoring object category.
4. The method according to claim 1 or 2, characterized in that the method further comprises: Acquiring a training monitoring object identity signal set, wherein one training monitoring object identity signal corresponds to one existing monitoring object category, and the training monitoring object identity signal set comprises a plurality of training monitoring object identity signals under each monitoring object category; Respectively extracting characteristics of each training monitoring object identity signal in the training monitoring object identity signal set to obtain training radio frequency fingerprint characteristics of each training monitoring object identity signal so as to obtain each training radio frequency fingerprint characteristic under each monitoring object category, wherein one training radio frequency fingerprint characteristic under one monitoring object category is the training radio frequency fingerprint characteristic of one training monitoring object identity signal belonging to the corresponding monitoring object category in the training monitoring object identity signal set; Determining an initial weight set under any monitoring object category in the monitoring object category set, and respectively calculating the category center distance of each training radio frequency fingerprint feature under any monitoring object category based on the category center of the any monitoring object category, each training radio frequency fingerprint feature under any monitoring object category and the initial weight set, wherein the initial weight set under any monitoring object comprises initial weight values of each distance calculation mode in a plurality of distance calculation modes under any monitoring object category; determining the class center distance of each screening radio frequency fingerprint feature in a plurality of screening radio frequency fingerprint features based on the class center distance of each training radio frequency fingerprint feature in any monitoring object class, and determining a target fitting distribution evaluation result based on the class center distance of each screening radio frequency fingerprint feature; And optimizing the initial weight set under any monitoring object category to minimize the target fitting distribution evaluation result until reaching a fitting convergence condition, thereby obtaining the target weight set under any monitoring object category, wherein the target weight set under any monitoring object category is used for calculating the category center distance of the target radio frequency fingerprint feature under any monitoring object category.
5. The method of claim 4, wherein the class center distance of one selected rf fingerprint is greater than the class center distance of any of the training rf fingerprints other than the plurality of selected rf fingerprints in all of the training rf fingerprints in any of the monitor classes, wherein the determining the target fit distribution evaluation result based on the class center distances of the selected rf fingerprints comprises: performing cauchy distribution fitting by adopting the class center distance of each screening radio frequency fingerprint feature in the plurality of screening radio frequency fingerprint features to obtain an initial cauchy distribution function under any monitoring object class; Determining a distance fitting value of each screening radio frequency fingerprint characteristic by adopting an initial Cauchy distribution function under any monitoring object category; And determining a target fitting distribution evaluation result based on the class center distance and the distance fitting value of each screening radio frequency fingerprint characteristic, wherein the target cauchy probability cumulative distribution under any monitoring object class is a cauchy probability cumulative distribution corresponding to the target weight group under any monitoring object class.
6. The method according to claim 1 or 2, wherein the target class probabilities of the target radio frequency fingerprint features under the respective monitoring object classes are determined by invoking a target normalization function, the method further comprising: Acquiring a test monitor identity signal set, wherein the test monitor identity signal set comprises a test class open set and a test class closed set, the test monitor identity signal in the test class open set belongs to an unknown class, and the test monitor identity signal in the test class closed set belongs to a known monitor class; Respectively extracting characteristics of each test monitoring object identity signal in the test monitoring object identity signal set to obtain test radio frequency fingerprint characteristics of each test monitoring object identity signal; Initializing a current normalization function, and determining the maximum confidence of each test monitoring object identity signal based on the test radio frequency fingerprint characteristics of each test monitoring object identity signal and the current normalization function, wherein one normalization function comprises a temperature parameter; Determining a maximum confidence coefficient statistical result of the test class open set and a maximum confidence coefficient statistical result of the test class closed set based on the maximum confidence coefficient of each test monitoring object identity signal, and judging whether a target probability distribution condition is met or not based on the maximum confidence coefficient statistical result of the test class open set and the maximum confidence coefficient statistical result of the test class closed set; and if the target probability distribution condition is not met, continuously optimizing the temperature parameter in the current normalization function until the target probability distribution condition is determined to be met.
7. The method of claim 6, wherein the target probability distribution condition is used to indicate that the probability distribution of the open set of test categories and the probability distribution of the closed set of test categories meet a separation requirement at a specified confidence threshold, wherein the determining whether the target probability distribution condition is met based on the maximum confidence statistics of the open set of test categories and the maximum confidence statistics of the closed set of test categories comprises: Based on the maximum confidence coefficient statistical result of the test class open set and the appointed confidence coefficient threshold value, determining open set probability distribution dividing information corresponding to the test class open set, wherein the open set probability distribution dividing information supports a relationship between the maximum confidence coefficient used for indicating the identity signal of the test monitoring object in the test class open set and the appointed confidence coefficient threshold value; Determining closed set probability distribution partition information corresponding to the closed set of test categories based on a maximum confidence coefficient statistical result of the closed set of test categories and the specified confidence coefficient threshold value, wherein the closed set probability distribution partition information supports a relationship between the maximum confidence coefficient used for indicating identity signals of the test monitoring objects in the closed set of test categories and the specified confidence coefficient threshold value; Based on the open set probability distribution dividing information corresponding to the open set of the test category and the closed set probability distribution dividing information corresponding to the closed set of the test category, judging whether a target probability distribution condition is met or not, so that the overlapping between the probability distribution of the open set of the test category and the probability distribution of the closed set of the test category is reduced through the target probability distribution condition.
8. An open set radio frequency fingerprint identification device, the device comprising: The acquisition unit is used for acquiring the identity signal of the target monitoring object; The processing unit is used for extracting the characteristics of the identity signal of the target monitoring object to obtain the target radio frequency fingerprint characteristics of the identity signal of the target monitoring object; The processing unit is further used for determining the class center of each monitoring object class in the monitoring object class set, and calculating the class center distance of the target radio frequency fingerprint feature under each monitoring object class based on the class center of each monitoring object class; The processing unit is further used for calculating tail probability of the target radio frequency fingerprint feature under each monitoring object category based on category center distance of the target radio frequency fingerprint feature under each monitoring object category and target cauchy probability cumulative distribution under each monitoring object category respectively, wherein the tail probability of the target radio frequency fingerprint feature under one monitoring object category is calculated by substituting the category center distance of the target radio frequency fingerprint feature under the corresponding monitoring object category into the target cauchy probability cumulative distribution under the corresponding monitoring object category; The processing unit is further configured to determine a target class probability of the target radio frequency fingerprint feature under each monitored object class by using a tail probability of the target radio frequency fingerprint feature under each monitored object class, and determine a radio frequency fingerprint identification result of the target radio frequency fingerprint feature based on the target class probability of the target radio frequency fingerprint feature under each monitored object class, where one radio frequency fingerprint identification result support is used for indicating whether the target radio frequency fingerprint feature is an existing monitored object class.
9. An electronic device, comprising: processor, and A memory in which a program is stored, Wherein the program comprises instructions which, when executed by the processor, cause the processor to perform the method according to any of claims 1-7.
10. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1-7.

Description

Open set radio frequency fingerprint identification method and device, storage medium and electronic equipment Technical Field The present invention relates to the field of radio frequency fingerprint identification technologies, and in particular, to an open set radio frequency fingerprint identification method, an open set radio frequency fingerprint identification device, a storage medium, and an electronic device. Background At present, the radio frequency fingerprint identification (Radio Frequency Fingerprint Identification, RFF) technology is used as a physical layer security means, and the identity of a radiation source is identified by extracting fine and unique physical characteristics introduced by transmitter hardware (such as a digital-to-analog converter, a mixer, a power amplifier and the like), so that the defect of protocol layer authentication can be effectively overcome. However, the related technology means is generally based on a closed set recognition assumption, that is, all signal categories occurring in a hypothesis test stage and the like occur in a training stage, so as to obtain higher recognition accuracy in such a closed set scene, but when an unknown category signal which is not included in training is encountered, a model can judge the unknown category signal as a known category, so that identity information recognition is invalid, and the accuracy of radio frequency fingerprint recognition is lower. Based on this, how to improve the accuracy of the radio frequency fingerprint identification has not yet had a better solution. Disclosure of Invention In view of the above, embodiments of the present invention provide an open-set radio frequency fingerprint identification method, apparatus, storage medium, and electronic device, so as to solve the problems of low accuracy of radio frequency fingerprint identification in the related art, that is, the embodiments of the present invention can determine, by using a category center distance of a target radio frequency fingerprint feature under each monitored object category and a target cauchy probability cumulative distribution under each monitored object category, a target category probability with high accuracy, so as to obtain a radio frequency fingerprint identification result with high accuracy, thereby effectively improving accuracy of radio frequency fingerprint identification, and further determining an unknown category when encountering an unknown monitored object category, that is, effectively improving perceptibility of the unknown category, so as to effectively avoid identity signal identification failure. According to an aspect of the present invention, there is provided an open set radio frequency fingerprint identification method, the method comprising: Acquiring a target monitoring object identity signal, and extracting characteristics of the target monitoring object identity signal to obtain a target radio frequency fingerprint characteristic of the target monitoring object identity signal; Determining class centers of all monitoring object classes in a monitoring object class set, and calculating class center distances of the target radio frequency fingerprint features under all the monitoring object classes based on the class centers of all the monitoring object classes respectively; Calculating the tail probability of the target radio frequency fingerprint feature under each monitoring object category based on the category center distance of the target radio frequency fingerprint feature under each monitoring object category and the target cauchy probability cumulative distribution under each monitoring object category, wherein the tail probability of the target radio frequency fingerprint feature under one monitoring object category is calculated by substituting the category center distance of the target radio frequency fingerprint feature under the corresponding monitoring object category into the target cauchy probability cumulative distribution under the corresponding monitoring object category; and determining a radio frequency fingerprint identification result of the target radio frequency fingerprint feature based on the target class probability of the target radio frequency fingerprint feature under each monitoring object class, wherein one radio frequency fingerprint identification result support is used for indicating whether the target radio frequency fingerprint feature is the existing monitoring object class or not. According to another aspect of the present invention, there is provided an open set radio frequency fingerprint identification device, the device comprising: The acquisition unit is used for acquiring the identity signal of the target monitoring object; The processing unit is used for extracting the characteristics of the identity signal of the target monitoring object to obtain the target radio frequency fingerprint characteristics of the identity signal of the target monitoring object;