A method for extracting radio frequency fingerprints of spread spectrum signals resistant to channel characteristic interference

Abstract

The present invention provides a method for extracting radio frequency fingerprints of spread spectrum signals based on resistance to channel characteristic interference, comprising the following steps: receiving a signal and performing down-conversion to obtain a baseband signal; pre-processing the baseband signal, extracting all leading code symbols from each frame signal, and performing frequency domain transformation on each leading code symbol; performing logarithmic operation on the frequency domain signal, performing centralization processing, and obtaining a complex inverse frequency domain signal; calculating the cyclic cross-correlation results of each complex inverse frequency domain signal in a frame signal and a reference complex inverse frequency domain signal, and performing filtering and smoothing processing; summing and averaging the cyclic cross-correlation results of each frame to obtain a radio frequency fingerprint. The present invention utilizes the physical layer characteristics of the signal as the radio frequency fingerprint of the device, separates the signal from the channel interference in the inverse spectrum domain, is applicable to spread spectrum signals with cyclic shifts, can realize identity authentication and device identification in multiple scenarios, eliminates the interference of channel characteristics, and improves the robustness of the radio frequency fingerprint.

Description

Technical Field

[0001] The present invention belongs to the field of communication and physical layer security technology, and in particular relates to a spread spectrum signal radio frequency fingerprint extraction method that is resistant to channel characteristic interference. Background Art

[0002] Radio frequency fingerprinting technology authenticates wireless devices at the physical layer using radio frequency signals. It is a crucial component for ensuring wireless network security and communications. Compared to traditional cryptographic methods, radio frequency fingerprinting offers a better approach to wireless device authentication. Radio frequency fingerprinting uses hardware defects embedded in the signals emitted by radio frequency devices to distinguish wireless devices. The power amplifier, mixer, and local crystal oscillator on the transmitter side of radio frequency devices all exhibit nonlinear deviations. These deviations are caused by the device's physical characteristics, the composition of electronic components, internal circuit board routing, and the device's usage and aging. Due to variations in manufacturing processes, radio frequency signals carry nonlinear hardware defects, which make it possible for receivers to distinguish transmitters. Even when manufactured using the same process, variations in radio frequency fingerprints can occur due to variations in production standards and tolerances.

[0003] The RF fingerprint features used in the RF fingerprint recognition method are divided into steady-state RF fingerprint features and transient RF fingerprint features. Steady-state RF fingerprint features are generated by the steady-state part of the signal, including carrier offset, IQ offset, differential constellation trajectory diagram, etc. Transient RF fingerprint features are generated by the state change of the transmitter, such as the pulse response when the device is turned on, including transient length, amplitude variance, number of peaks of the carrier signal, and the difference between the normalized average and normalized maximum values ​​of the transient power. This transition from off to on occurs before the actual transmission of the signal, and the starting point of the transient signal needs to be accurately extracted. The current RF fingerprint recognition method is susceptible to environmental changes and equipment aging. In order to resist noise interference, the communication distance is shortened, and the tiny RF fingerprint is easily distorted by channel interference, which reduces the accuracy of recognition.

[0004] In wireless communications, wireless channel variations are crucial. The wireless channel and the device's transmitted signal manifest as linear convolution in the time domain, multiplication in the frequency domain, and addition in the cepstrum. Due to the influence of wireless channel characteristics, device fingerprints can also be distorted to some extent. In real-world communication environments, channel interference far outweighs variations in the device's fingerprint, and the resulting signal distortion is difficult to address in the time domain. Therefore, cepstrum separation is introduced to identify the channel and transmitted signal. The cyclic cross-correlation function is used to remove weakly correlated components in different scenarios, resulting in a purer signal and facilitating RF fingerprint extraction. Summary of the Invention

[0005] To overcome the shortcomings of the existing technology, the present invention provides a method for extracting radio frequency fingerprints from spread-spectrum signals that is resistant to channel interference. This method uses the cyclic cross-correlation results of physical-layer features as the device's radio frequency fingerprint, increasing the dimensionality and data volume of the signal fingerprint. Furthermore, the identical cross-power spectra between cyclically shifted signals are exploited to achieve data independence of the radio frequency fingerprint. This method leverages the stability of frequency-domain features to resist environmental noise, making it suitable for low signal-to-noise ratio environments and meeting the requirements of long-distance communication.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A method for extracting radio frequency fingerprints of spread spectrum signals that is resistant to channel characteristic interference comprises the following steps:

[0008] Step 1: Receive the signal: After receiving the signal, perform down-conversion processing to obtain the baseband signal; read the baseband signal and normalize it, perform energy retrieval to separate the effective signal segment;

[0009] Step 2, data preprocessing: perform a cross-correlation operation on each signal segment with the local reference preamble signal, obtain the data point with the largest peak value as the preamble starting data point, and perform data preprocessing on each preamble symbol;

[0010] Step 3, inverse frequency domain transformation: Perform frequency domain transformation on each preamble symbol in a frame signal, and then perform logarithmic operation and centralization to obtain a complex inverse frequency domain signal;

[0011] Step 4, cross-correlation operation: Calculate the cyclic cross-correlation result of the complex inverse frequency domain signal of each preamble symbol and the complex inverse frequency domain signal of the reference symbol, perform filtering and smoothing processing, and sum and average to obtain the cyclic cross-correlation result of one frame of signal;

[0012] Step 5, feature extraction: sum and average the cyclic cross-correlation results of each frame signal to obtain the RF fingerprint.

[0013] Furthermore, in step 1, the process of performing energy retrieval to separate the valid signal segments includes:

[0014] Use the sliding window method to perform energy retrieval on the data segment and calculate the total energy within a window:

[0015]

[0016] Among them, l w =0,1,…,L w -1, L w is the sliding window length; r n is the amplitude value of the nth received signal point, E n The length is L wThe sliding window of the sum of the energy values ​​of all signal points starting from the nth received signal point; the window is slid by adding the previous point of the window and removing the last point of the window, so the sum of the energy values ​​of the next window is:

[0017]

[0018] When the total energy of the window exceeds the threshold, it is considered that a valid signal is successfully found.

[0019] Furthermore, in step 2, data preprocessing includes: coarse frequency offset estimation, fine frequency offset estimation, time synchronization, symbol synchronization, carrier synchronization and data normalization processing.

[0020] Furthermore, the step 3 specifically includes the following process:

[0021] Perform Fourier transform on each preamble symbol:

[0022] Y i,j (k) = FFT(y i,j (n))

[0023] Where n = 0, 1, ..., M-1, k = 0, 1, ..., M-1, M is the number of sampling points of a symbol; i = 0, 1, ..., L-1, L is the number of collected signal frames; j = 0, 1, ..., N-1, N is the number of preamble symbols contained in a frame of signal; Y i,j (k) is the frequency domain signal of the jth preamble symbol of the i-th segment signal, y i,j (n) is the jth preamble symbol of the i-th segment signal, and FFT() is the Fourier transform function;

[0024] Then perform logarithmic operation on each frequency domain signal:

[0025]

[0026] in, is the complex inverse frequency domain signal of the j-th preamble symbol of the i-th segment signal;

[0027] Then each complex inverse frequency domain signal is centrally processed:

[0028]

[0029] in, is the complex inverse frequency domain signal of the j-th preamble symbol of the i-th segment signal after zero-meaning, and M is the number of sampling points of one symbol.

[0030] Furthermore, the step 4 specifically includes the following process:

[0031] Calculate the complex inverse frequency domain signal for each symbol The complex inverse frequency domain signal X of the local reference symbol i,j The cyclic cross-correlation function of (k) is:

[0032]

[0033] Where, k = 0, 1, ..., M-1, M is the number of sampling points of a symbol; R i,j (k) is the cyclic cross-correlation result calculated for the j-th preamble symbol of the i-th segment signal, and ⊙ is the cyclic cross-correlation operation;

[0034] Then the cyclic cross-correlation result R i,j (k) Perform smoothing filtering, and select the filter as follows:

[0035]

[0036] in, The value of m is determined by the low-frequency part of the signal; the cyclic cross-correlation results of all preamble symbols in a frame of signal are windowed and summed up to be:

[0037]

[0038] Wherein, N is the number of preamble symbols contained in a frame signal.

[0039] Furthermore, the specific calculation formula in step 5 is:

[0040]

[0041] Where L is the number of collected signal frames.

[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0043] (1) The method of identity authentication and device identification based on physical layer characteristics has good security, uniqueness and difficulty in tampering. The cyclic cross-correlation result is used as the device radio frequency fingerprint, which can resist additive noise in the environment and has good stability.

[0044] (2) Using the cepstrum to separate the signal, channel, and environmental interference in the frequency domain, the transformed signal energy is more concentrated, and the accuracy of the re-transformation is improved. This makes it easier to separate and identify the signal, extract a purer signal, and better reveal the essential characteristics of the device fingerprint.

[0045] (3) Utilize the cyclic shift characteristics of the spread spectrum signal to achieve data independence of the RF fingerprint. Using the cyclic cross-correlation function to remove irrelevant components in different scenarios, including channel, environment, and multipath effects, the relevant components between different signals are separated, namely the device RF fingerprint. This can achieve identity authentication and device identification in multiple scenarios, better resist channel characteristic interference, and improve the robustness of the RF fingerprint. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of the method for extracting radio frequency fingerprints of spread spectrum signals that is resistant to channel characteristic interference according to the present invention;

[0047] Figure 2 This is a schematic diagram of the frame structure format of the physical layer signal frame of the LoRa device in the present invention;

[0048] Figure 3 Schematic diagram of the complex inverted frequency domain signal of the preamble symbol in the present invention;

[0049] Figure 4 Schematic diagram of the complex cepstrum domain cyclic cross-correlation result between the preamble symbol and the local reference symbol in the present invention;

[0050] Figure 5 Schematic diagram of the radio frequency fingerprint of the LoRa device in the present invention. DETAILED DESCRIPTION

[0051] The technical solutions provided by the present invention will be described in detail below with reference to specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0052] The present invention proposes a method for extracting radio frequency fingerprints of spread spectrum signals that is resistant to channel characteristic interference. The overall flow chart is as follows: Figure 1 As shown, the specific steps include:

[0053] Step 1, receiving the signal: after receiving the signal, down-convert it to obtain the baseband signal. The demodulated binary baseband LoRa signal is read in the format of complex64, that is, a complex number with 32 bits for both real and imaginary parts. The LoRa signal is composed of chirp signals, such as Figure 2 As shown in Figure 1, its frame structure consists of three parts: preamble, header, and data payload. The preamble consists of 8 up-chirp signals, 2 frame synchronization words, and 2.25 down-chirp signals. The remaining part is the result of cyclic shift of the reference up-chirp signal. After normalizing the data, the sliding window method is used to perform energy retrieval on the data segment. By calculating the total energy within a window, the influence of occasional high-intensity noise on the total energy of the signal within the window is eliminated. The energy of a sliding window with a length of 200 is:

[0054]

[0055] Among them, l w =0,1,…,199,l w is the sampling point in the sliding window; r n is the amplitude value of the nth received signal point, E n is the sum of the energy values ​​of all signal points starting from the nth received signal point in a sliding window of length 200. The window is slid by adding the previous point in the window and removing the last point in the window, so the sum of the energy values ​​of the next window is:

[0056] E n+1 =E n +|r n+1 | 2 -|r n-199 |2

[0057] When the total window energy exceeds the threshold of 0.05, it is considered that a valid signal is successfully found.

[0058] Step 2, data preprocessing: perform cross-correlation operation on each signal segment with a local single reference up-chirp signal. Each chirp signal contains 1024 sampling points. Let the received signal with a length of 1024 sample points be y r (n), the local single reference up-chirp signal is x(n), and the complex conjugate correlation value between the two is calculated as:

[0059]

[0060] Among them, x * (n) is the conjugate of x(n). Set the threshold value TH to 100. When r(n)>TH, it is considered that a preamble symbol is captured. After capturing 8 preamble symbols continuously, after expanding 5 sampling points before and after, sliding correlation is performed with the local standard preamble signal segment. The sampling point with the largest cross-correlation value is selected as the preamble starting point, and the preamble symbol of the captured i-th frame signal is separated as Where i = 0, 1,…, 99, j = 0, 1,…, 7, n = 0, 1,…, 1023.

[0061] Then, the frequency offset of the leading code symbol of a frame signal is estimated and compensated. The average phase difference between the corresponding points m of the adjacent symbols is calculated as:

[0062]

[0063] Where l = 0, 1, ..., 7, so the average phase deviation of the preamble is:

[0064]

[0065] Therefore, the average frequency deviation of the preamble part is expressed as:

[0066]

[0067] Among them, the sampling frequency f s For the signal after symbol synchronization Perform frequency offset correction and phase offset correction. Then normalize the 8 preamble codes and obtain the signal after carrier synchronization:

[0068]

[0069] Among them, n=0,1,…,1023, i=0,1,…,99, j=0,1,…,7.

[0070] Step 3, inverse frequency domain transform: for each preamble symbol y i,j (n) Perform Fourier transform processing respectively:

[0071] Y i,j (k) = FFT(y i,j (n))

[0072] Where n = 0, 1, ..., 1023, k = 0, 1, ..., 1023, i = 0, 1, ..., 99, j = 0, 1, ..., 7. i,j (k) is the frequency domain signal of the jth preamble symbol of the i-th segment signal, y i,j (n) is the j-th preamble symbol of the i-th segment signal, and FFT() is the Fourier transform function.

[0073] Then for each frequency domain signal Y i,j (k) Perform logarithmic operation:

[0074]

[0075] in, is the complex inverse frequency domain signal of the j-th preamble symbol of the i-th segment signal.

[0076] Then, each complex inverse frequency domain signal is centrally processed, that is, the mean of each signal is removed:

[0077]

[0078] in, is the complex inverse frequency domain signal after zero-meaning of the j-th preamble symbol of the i-th segment signal, and its schematic diagram is as follows Figure 3 shown.

[0079] Step 4, cross-correlation operation: Calculate the complex inverse frequency domain signal of each symbol The complex inverse frequency domain signal X of the local reference symbol i,j (k) the cyclic cross-correlation function R i,j (k) is:

[0080]

[0081] Where k = 0, 1, ..., 1023. i,j (k) is the cyclic cross-correlation result calculated for the j-th preamble symbol of the i-th segment signal, and ⊙ is the cyclic cross-correlation operation.

[0082] Since the signal is concentrated in the low frequency part, while the noise, environment and other interferences are concentrated in the high frequency part. Therefore, the cyclic cross-correlation result R i,j (k) is smoothed and filtered, and the low-frequency signal segments of ten sampling points before and after are superimposed to remove high-frequency burrs. Therefore, the filter W(k) is selected as:

[0083]

[0084] The cyclic cross-correlation results of the eight preamble symbols in a frame of signal are windowed and summed up to be equal:

[0085]

[0086] Among them, RFF i (k) is the RF fingerprint of the i-th segment signal, and its schematic diagram is as follows Figure 4 shown.

[0087] Step 5, feature extraction: sum and average the cyclic cross-correlation results of 100 frames of device signals and local reference signals to obtain Figure 5 The RF fingerprint shown:

[0088]

[0089] The technical means disclosed in the solutions of the present invention are not limited to those disclosed in the above-mentioned embodiments, but also include technical solutions composed of any combination of the above-mentioned technical features. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for extracting radio frequency fingerprints of spread spectrum signals that is resistant to channel characteristic interference, characterized in that: The following steps are involved: Step 1: Receive the signal: After receiving the signal, perform down-conversion processing to obtain the baseband signal; read the baseband signal and normalize it, perform energy retrieval to separate the effective signal segment; Step 2, data preprocessing: perform a cross-correlation operation on each signal segment with the local reference preamble signal, obtain the data point with the largest peak value as the preamble starting data point, and perform data preprocessing on each preamble symbol; Step 3, inverse frequency domain transformation: Perform frequency domain transformation on each preamble symbol in a frame signal, and then perform logarithmic operation and centralization to obtain a complex inverse frequency domain signal; Step 4, cross-correlation operation: Calculate the cyclic cross-correlation result of the complex inverse frequency domain signal of each preamble symbol and the complex inverse frequency domain signal of the reference symbol, perform filtering and smoothing processing, and sum and average to obtain the cyclic cross-correlation result of one frame of signal; Step 5, feature extraction: sum and average the cyclic cross-correlation results of each frame signal to obtain the RF fingerprint.

2. The method for extracting radio frequency fingerprints of spread spectrum signals resistant to channel characteristic interference according to claim 1, characterized in that: In step 1, the process of performing energy retrieval to separate the effective signal segments includes: Use the sliding window method to perform energy retrieval on the data segment and calculate the total energy within a window: Among them, l w =0,1,…,L w -1, L w is the sliding window length; r n is the amplitude value of the nth received signal point, E n The length is L w The sliding window of the sum of the energy values ​​of all signal points starting from the nth received signal point; the window is slid by adding the previous point of the window and removing the last point of the window, so the sum of the energy values ​​of the next window is: When the total energy of the window exceeds the threshold, it is considered that a valid signal is successfully found.

3. The method for extracting radio frequency fingerprint of spread spectrum signals resistant to channel characteristic interference according to claim 1, characterized in that: In step 2, data preprocessing includes: coarse frequency offset estimation, fine frequency offset estimation, time synchronization, symbol synchronization, carrier synchronization and data normalization processing.

4. The method for extracting radio frequency fingerprints of spread spectrum signals resistant to channel characteristic interference according to claim 1, characterized in that: The step 3 specifically includes the following process: Perform Fourier transform on each preamble symbol: Y i,j (k)=FFT(y i，j (n)) Where n = 0, 1, ..., M-1, k = 0, 1, ..., M-1, M is the number of sampling points of a symbol; i = 0, 1, ..., L-1, L is the number of collected signal frames; j = 0, 1, ..., N-1, N is the number of preamble symbols contained in a frame of signal; Y i,j (k) is the frequency domain signal of the jth preamble symbol of the i-th segment signal, y i,j (n) is the jth preamble symbol of the i-th segment signal, and FFT() is the Fourier transform function; Then perform logarithmic operation on each frequency domain signal: in, is the complex inverse frequency domain signal of the j-th preamble symbol of the i-th segment signal; Then each complex inverse frequency domain signal is centrally processed: in, is the complex inverse frequency domain signal of the j-th preamble symbol of the i-th segment signal after zero-meaning, and M is the number of sampling points of one symbol.

5. The method for extracting radio frequency fingerprint of spread spectrum signals resistant to channel characteristic interference according to claim 1, characterized in that: The step 4 specifically includes the following process: Calculate the complex inverse frequency domain signal for each symbol The complex inverse frequency domain signal X of the local reference symbol i,j The cyclic cross-correlation function of (k) is: Where, k = 0, 1, ..., M-1, M is the number of sampling points of a symbol; R i,j (k) is the cyclic cross-correlation result calculated for the j-th preamble symbol of the i-th segment signal, and ⊙ is the cyclic cross-correlation operation; Then the cyclic cross-correlation result R i,j (k) Perform smoothing filtering, and select the filter as follows: in, The value of m is determined by the low-frequency part of the signal; the cyclic cross-correlation results of all preamble symbols in a frame of signal are windowed and summed up to be: Wherein, N is the number of preamble symbols contained in a frame signal.

6. The method for extracting radio frequency fingerprints of spread spectrum signals resistant to channel characteristic interference according to claim 1, characterized in that: The specific calculation formula in step 5 is: Where L is the number of collected signal frames.

Images