A 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint

Abstract

The present invention discloses a 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint. The method uses a universal software radio platform USRP to collect steady-state radio frequency signals of device communication, performs subsequent signal processing and feature extraction processing, and finally realizes device classification. Compared with the existing technology, this method uses a universal software radio platform to collect signals at a lower sampling rate, which is low-cost; the identity authentication means at the software level can be tampered with and forged by malicious users, thereby evading authentication. This method uses the inherent physical defects of the radio frequency transceiver of the 5G terminal device to form a device fingerprint, which is an efficient and low-cost device authentication solution. This method proposes a method based on preamble code threshold detection to determine the fingerprint signal segment of the device. After verification by a large number of device experiments, the statistical authentication accuracy can be as high as 95.6%.

Description

Technical Field

[0001] The present invention belongs to the field of mobile communications and relates to a 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint. Background Art

[0002] In recent years, mobile communications have seen exponential growth in both the number of devices and the amount of data being transmitted, with devices becoming increasingly complex and intelligent. With the continuous maturation of computer network and communication technologies, wireless networks, with their advantages of ease of migration, rapid networking, and convenient installation, have provided significant benefits to human society. In today's rapidly developing information society, information security is paramount. However, the openness, mobility, and dynamically changing topology of wireless networks make them vulnerable to a variety of malicious attacks, from passive to active, posing a serious challenge to information security. Wireless network security issues cannot be ignored. Due to the limited battery life and computing power of wireless mobile devices, cyberattacks pose a potential threat, and the number of attacks is growing exponentially. According to reports, the current cost of cybercrime has exceeded $600 billion. In traditional wireless communication systems, network access authentication is typically performed above the physical layer, primarily relying on stored device identity information (such as MAC addresses, SIM cards, security certificates, etc.) or input authentication passwords (such as WEP / WPA passwords and PIN codes) to prevent unauthorized access. Due to limited defenses against malicious attacks, devices rely on software-based authentication for identification. However, this identity can be tampered with by malicious users, circumventing the authentication mechanism. Malicious users can also forge identities to launch denial-of-service (DoS) attacks. Traditional password-based security methods are expensive to set up and complex and vulnerable to key distribution and management. Therefore, it is imperative to find a new terminal device authentication method that is robust and resistant to tampering and forgery, suitable for device authentication in mobile communication networks. Summary of the Invention

[0003] The present invention provides a 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint. By collecting the radio frequency signal during device communication, the device fingerprint caused by the inherent physical defects of the device itself is extracted and used for the identity authentication of the terminal device. It is independent of the identity authentication at the software level, thereby improving the reliability and security of the identity authentication.

[0004] The 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint of the present invention comprises the following steps:

[0005] 1) The 5G terminal device actively initiates a call and transmits a radio frequency signal.

[0006] 2) The universal software radio platform USRP is used to passively collect the RF signal of the 5G terminal device. The sampling frequency of the collected signal is 20MS / s. The I / Q data streams of the in-phase and quadrature phases are saved in the form of 8-bit integers. The obtained I / Q binary files are stored in the form of [I0, Q0, I1, Q1, ... I i ,Q i ,...I n ,Q n ] format interleaved stack storage, where n is the number of collected signal sample pairs, I i Represents the I data of the i-th group of signal samples, Q i It represents the Q data of the i-th group of signal samples. The collected data is processed according to the complex format, and the original data becomes [I0+jQ0,I1+jQ1,...I i +jQ i ,...I n +jQ n ] format, where j is the imaginary basic unit in the complex signal domain.

[0007] 3) Use the preamble threshold detection algorithm to process the collected signal samples, which specifically includes the following four steps:

[0008] (1) Evaluate the channel noise by scanning at least 1000 samples to obtain the maximum noise value S noise ;

[0009] (2) According to the maximum noise value obtained in step (1), the preamble threshold is set to S th =S noise +0.1;

[0010] (3) Scan the signal and select the first signal that is higher than the threshold S th =S noise The sample value of +0.1 is saved as the preamble start index h;

[0011] (4) Starting from the preamble start index, collect the signal with a duration of 20ms and intercept the signal segment [h, h+20ms].

[0012] 4) Repeat steps 1)-3) to obtain several signals processed by the preamble threshold detection algorithm;

[0013] 5) Performing time domain processing on the signal samples collected in step 4) to obtain three variables: instantaneous amplitude (IA), instantaneous phase (IP), and instantaneous frequency (IF), and performing normalization on the processed variables.

[0014] 6) Calculate the variance, skewness, kurtosis, and standard deviation of several normalized signal samples (the signal samples are continuous signal segments of equal length covering the entire preamble area) to form a feature vector. The dimension of the feature vector is N features ×N subregions ×N domains =4×29×3=348, where N features Refers to the number of features that make up the feature vector, N subregions Refers to the number of consecutive signal segments of equal length that cover the entire preamble area, N domains Refers to the number of variables obtained after time domain processing of signal samples.

[0015] 7) Use a supervised learning method to train the feature vector to obtain a radio frequency fingerprint authentication model, and implement device authentication based on the radio frequency fingerprint authentication model. Use the trained radio frequency fingerprint authentication model to authenticate the 5G terminal device to be authenticated, set a threshold θ = 0.6, and if the similarity between the feature vector of the device to be authenticated and the existing feature vector is greater than the set threshold θ = 0.6, then the device is judged to have passed the authentication; otherwise, it is judged to have failed the authentication.

[0016] To evaluate the system effect, use the formula Calculate the accuracy, where TP refers to the number of samples correctly detected as positive, TN refers to the number of samples correctly detected as negative, FP refers to the number of samples falsely detected as positive, and FN refers to the number of samples falsely detected as negative.

[0017] The beneficial effects of the present invention are:

[0018] The method of the present invention provides a method for device detection and authentication in 5G mobile networks using radio frequency fingerprints. It uses a universal software radio platform (USRP) to collect steady-state radio frequency signals from device communications, employs a preamble threshold detection algorithm to intercept preamble signal segments, and performs subsequent signal processing and feature extraction, ultimately achieving device classification and authentication. Compared with existing technologies, the universal software radio platform (USRP) used in the method of the present invention can collect signals at a lower sampling rate, resulting in lower costs and simpler system construction. The method of the present invention avoids the problem that software-level identity authentication methods can be tampered with and forged by malicious users. It utilizes the inherent physical defects of the RF transceiver of 5G terminal devices to construct a device fingerprint. This characteristic fingerprint is unique, robust, long-term invariant, independent, and unified, ensuring high efficiency while minimizing system construction and maintenance costs. The method proposed in the present invention, based on preamble threshold detection, determines fingerprint signal segments. After verification in a large number of device experiments, the statistically obtained authentication accuracy rate can reach up to 95.6%. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be further described below with reference to the embodiments and accompanying drawings.

[0021] like Figure 1 Shown is the method flow of an embodiment of the present invention.

[0022] A 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint includes the following steps:

[0023] 1) Deploy a general software radio platform at a fixed distance from the device to be collected, operate the 5G terminal device to actively call and send out RF signals, and at the same time, the software radio platform collects signals at a fixed sampling rate and fixed duration. The sampling frequency of the collected signals is 20MS / s, and the in-phase and quadrature-phase I / Q data streams are saved as 8-bit integers respectively. The sampling time is set to 5s.

[0024] 2) First, sample the channel 1000 times to determine the maximum noise value S of the channel noise , then set the preamble threshold to S th =S noise +0.1, scan the signal and select the first signal above the threshold S th =S noise The sample value of +0.1 is saved as the preamble code starting index h. Starting from the preamble code starting index, the preamble code signal segment is intercepted. The intercepted signal duration is 20ms, and finally the preamble code signal segment [h, h+20ms] with a duration of 20ms is obtained.

[0025] 3) Repeat steps 1) and 2) multiple times to increase the number of samples to ensure the robustness of the acquisition.

[0026] 4) Perform time domain processing on the intercepted preamble code segment to obtain instantaneous amplitude (IA), instantaneous phase (IP), and instantaneous frequency (IF) responses, and normalize the responses.

[0027] 5) For N r = 29 continuous signal segments of equal length covering the entire preamble area are calculated to obtain the variance (Variance), skewness (Skewness), kurtosis (Kurtosis) and standard deviation (Standard Deviation) of the processed signal samples (the calculation formula is shown in Table 1), which are used to form a feature vector. The dimension of the feature vector is calculated as: N features ×N subregions ×N domains =4×29×3=348, where N features Refers to the number of features that make up the feature vector, N subregions Refers to the number of consecutive signal segments of equal length that cover the entire preamble area, N domains Refers to the number of variables obtained after time domain processing of signal samples.

[0028] 6) Use a supervised learning method to train the feature vector to obtain a radio frequency fingerprint authentication model, and implement device authentication based on the radio frequency fingerprint authentication model. Use the trained radio frequency fingerprint authentication model to authenticate the 5G terminal device to be authenticated, set a threshold θ = 0.6, and if the similarity between the feature vector of the device to be authenticated and the existing feature vector is greater than the set threshold θ = 0.6, the device is judged to have passed the authentication, otherwise it is judged to have failed the authentication.

[0029] Table 1 Calculation formulas for the four features

[0030]

Claims

1. A 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint, characterized in that: The method comprises the following steps: 1) The 5G terminal device actively initiates a call and transmits a radio frequency signal; 2) Use the Universal Software Radio Platform (USRP) to passively collect RF signals transmitted by 5G terminal devices; 3) Processing the collected signal samples using a preamble threshold detection algorithm; the preamble threshold detection algorithm includes the following steps: (1) Evaluate the channel noise by scanning at least 1000 samples to obtain the maximum noise value ; (2) According to the maximum noise value obtained in step (1), the preamble threshold is set to ; (3) Scan the signal and select the first signal that is higher than the threshold The sample value is saved as the preamble start index h; (4) Starting from the preamble code start index, collect the signal for 20ms and intercept the signal segment ; 4) Repeat steps 1)-3) to obtain several signals processed by the preamble threshold detection algorithm; 5) Perform time domain processing on the signal samples collected in step 4) to obtain three variables: instantaneous amplitude, instantaneous phase, and instantaneous frequency, and perform normalization on the processed variables; 6) Calculate the variance, skewness, kurtosis, and standard deviation of the normalized signal sample to form a feature vector; 7) Using a supervised learning method to train the feature vector to obtain a radio frequency fingerprint authentication model, and implementing device authentication based on the radio frequency fingerprint authentication model.

2. The 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint according to claim 1 is characterized in that: In step 2), the sampling frequency of the collected signal is 20MS / s, and the I / Q data streams of the in-phase and quadrature phases are saved in the form of 8-bit integers. The obtained I / Q binary files are stored as follows: Format interleaved stack storage, where n is the number of collected signal sample pairs, represents the I data of the i-th group of signal samples, Represents the Q data of the i-th group of signal samples. The collected data is processed according to the complex format, and the original data becomes format, where j is the imaginary basic unit in the complex signal domain.

3. The 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint according to claim 1 is characterized in that: In step 6), the variance, skewness, kurtosis, and standard deviation of the signal samples are calculated for a number of continuous signal segments of equal length covering the entire preamble region.

4. The 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint according to claim 3 is characterized in that: In step 6), the dimension of the feature vector of each signal sample is ,in refers to the number of features that make up the eigenvector, Refers to the number of consecutive signal segments of equal length that cover the entire preamble area. Refers to the number of variables obtained after time domain processing of signal samples.

5. The 5G terminal device fingerprint extraction and authentication method based on radio frequency fingerprint according to claim 1 is characterized in that: In step 7), the trained RF fingerprint authentication model is used to authenticate the 5G terminal device to be authenticated, and a threshold is set. If the similarity between the feature vector of the device to be authenticated and the existing feature vector is greater than the set threshold , then the device is judged to have passed the authentication, otherwise it is judged to have failed the authentication.

Images