A radio frequency fingerprint extraction method based on LTE PRACH signal cepstrum and cross-correlation calculation

Abstract

The present invention relates to a radio frequency fingerprint extraction method based on LTE PRACH signal cepstrum and cross-correlation operation, the method comprising the following steps: step 1: sampling the LTE PRACH signal, step 2: synchronously processing the sampled signal, step 3: estimating and correcting the carrier frequency offset, step 4: performing fast Fourier transform on the signal after frequency offset removal, and extracting N ZC Valid subcarrier data, get the frequency domain ZC sequence, step 5: take the logarithm of the frequency domain ZC sequence obtained in step 4, get the cepstrum domain signal, step 6: perform cross-correlation operation on the first #imgabs0# signals and the last #imgabs1# signals in the cepstrum domain, and use the result as the RF fingerprint. This technical solution effectively separates the stable RF fingerprint component and the random channel fading component in the LTE received signal, and removes the random channel component from the received signal, thereby realizing the accurate extraction of the LTE device fingerprint and further improving the accuracy of the physical layer authentication of the LTE device.

Description

Technical Field

[0001] The invention relates to an extraction method, in particular to a radio frequency fingerprint extraction method based on LTE PRACH signal cepstrum and cross-correlation operations, and belongs to the technical field of wireless communication physical layer security. Background Art

[0002] The existing physical layer RF fingerprint extraction and authentication technologies for LTE uplink devices mainly include the differential trajectory constellation diagram method and the I / Q branch imbalance extraction method. The above methods do not effectively process the random channel components in the received signal. The random channel components cause a certain degree of disturbance in the extracted fingerprint, resulting in the inability to effectively extract the stable RF fingerprint components contained in the LTE uplink sampling signal, limiting the accuracy of LTE device identification and authentication. In addition, the above existing RF fingerprint extraction technology relies on local a priori PRACH sequence parameters, and the upper-layer system information module needs to be decoded before the RF fingerprint extraction, which increases the overall delay of identity authentication and reduces the actual availability. Therefore, there is an urgent need for a new solution to solve the above technical problems. Summary of the invention

[0003] The present invention aims at the problems existing in the prior art and provides a method for extracting radio frequency fingerprints based on LTE PRACH signal cepstrum and cross-correlation operations. The technical problem to be solved by this technical solution is to effectively separate the stable radio frequency fingerprint components and random channel fading components in the LTE received signal, and remove the random channel components from the received signal. This enables accurate extraction of LTE device fingerprints and further improves the accuracy of LTE device physical layer authentication.

[0004] In order to achieve the above object, the technical solution of the present invention is as follows: a radio frequency fingerprint extraction method based on LTE PRACH signal cepstrum and cross-correlation operation, the method comprising the following steps:

[0005] The present invention uses the cepstrum and cross-correlation calculation technology in signal processing to extract the radio frequency fingerprint of the LTE random access PRACH signal and identify the legitimacy of the device identity. The specific technical steps are as follows.

[0006] Step 1: Sample the LTE PRACH signal to obtain r(n).

[0007] Step 2: Perform synchronization preprocessing according to the following steps. Based on the cyclic prefix structure of the PRACH signal frame, the following formula can be used to achieve symbol synchronization in time:

[0008]

[0009] Where r(n) is the sampled PRACH signal, NCP is the sample length of the cyclic prefix, N SEQ is the sample length of the preamble sequence part in the PRACH frame, and N PRACH is the total sample length of the PRACH frame. The s value obtained according to formula (1) is the first sample point of the actual valid data.

[0010] Step 3: Carrier frequency offset estimation and correction. Using the cyclic prefix structure of the PRACH signal frame, the phase deviation of the signal sampling is first estimated as

[0011]

[0012] Where k = 1L L and L is the number of repeated PRACH frames. Then, the carrier frequency offset can be calculated as

[0013]

[0014] where f s is the sampling rate. The signal sampling carrier frequency offset compensation can be

[0015]

[0016] Step 4: Extract the ZC sequence in the frequency domain. Perform the CP removal operation on the synchronized and frequency-debiased PRACH signal frame to obtain the leading sequence part, which is expressed as

[0017]

[0018] Where s(n) is the PRACH signal at the transmitter, h(n) and η(n) are the channel and noise, represents convolution. Performing Fast Fourier Transform (FFT) on r(n) yields

[0019] R(ω)=W N [FFT(r(n))] = W N [(S(ω)+σ'(ω))H(w)] (6)

[0020] in W N [g] represents a window of length N. Therefore, through FFT transformation, the convolution channel effect in equation (5) is transformed into a multiplication effect.

[0021] Step 5: Further transform the multiplicative channel effect into additive effect through cepstrum operation

[0022] C(ω)=log[R(ω)]=W N [log[S(ω)+σ'(ω)]+log[H(ω)]] (7)

[0023] Note that the channel fading H(ω) at different frequencies ω is different and random.

[0024] Step 6: Further perform cross-correlation operation on the above cepstrum

[0025]

[0026] in Indicates rounding down, N ZC Indicates the length of the frequency domain ZC sequence.

[0027] Compared with the prior art, the present invention has the following advantages: the radio frequency fingerprint of the LTEPRACH signal obtained by the above method is sent to the deep learning network for training and classification, and the classification accuracy in the 2-meter line-of-sight (LOS) environment, the 6-meter line-of-sight (LOS) environment and the 4-meter non-line-of-sight (NLOS) environment can reach more than 99%, indicating that the present method can effectively combat the adverse effects of the wireless channel and extract a stable radio frequency fingerprint. The present invention makes full use of the constant amplitude characteristics of the LTE PRACH signal in the frequency domain, and the frequency domain ZC sequence provides a better unified representation for the radio frequency fingerprint; the random channel components in the PRACH signal are effectively separated from the stable radio frequency fingerprint components by using the cepstrum operation, and the cross-correlation operation removes the above random channel components, thereby increasing the stability of the extracted radio frequency fingerprint. In addition, the present invention does not need to decode the parameters of the upper-layer system information module, and gets rid of the dependence on prior parameter information in the process of extracting the radio frequency fingerprint, thereby reducing the implementation complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 Classification confusion matrix; (a) 2m line-of-sight (LOS) environment, (b) 6m line-of-sight (LOS) environment, (c) 4m non-line-of-sight (NLOS) environment;

[0029] Figure 2 Classification accuracy curve;

[0030] Figure 3 Classification accuracy histogram. DETAILED DESCRIPTION

[0031] In order to deepen the understanding of the present invention, the present embodiment is described in detail below with reference to the accompanying drawings.

[0032] Example 1: See Figure 1-Figure 3 , a radio frequency fingerprint extraction method based on LTE PRACH signal cepstrum and cross-correlation operation, the method comprises the following steps: using general software radio equipment to conduct experiments and obtain the results as shown in Figure 1 , Figure 2 and Figure 3 As shown. Figure 1 is the classification confusion matrix, Figure 2 is the classification accuracy curve of this method, and Figure 3 The classification accuracy histogram is compared with the traditional method (DCTF). The present invention uses the cepstrum and cross-correlation operation technology in signal processing to extract the radio frequency fingerprint of the LTE random access PRACH signal and identify the legitimacy of the device identity. The specific technical steps are as follows.

[0033] Step 1: Sample the LTE PRACH signal to obtain r(n);

[0034] Step 2: Perform synchronization preprocessing according to the following steps. Based on the cyclic prefix structure of the PRACH signal frame, the following formula can be used to achieve symbol synchronization in time:

[0035]

[0036] Where r(n) is the sampled PRACH signal, N CP is the sample length of the cyclic prefix, N SEQ is the sample length of the preamble sequence part in the PRACH frame, and N PRACH is the total sample length of the PRACH frame. The s value obtained according to formula (1) is the first sample point of the actual valid data.

[0037] Step 3: Carrier frequency offset estimation and correction. Using the cyclic prefix structure of the PRACH signal frame, the phase deviation of the signal sampling is first estimated as

[0038]

[0039] Where k = 1L L and L is the number of repeated PRACH frames. Then, the carrier frequency offset can be calculated as

[0040]

[0041] where f s is the sampling rate. The signal sampling carrier frequency offset compensation can be

[0042]

[0043] Since the PRACH signal uses the Zadoff-Chu (ZC) sequence, the good properties of ZC can be used to extract and process the RF fingerprint. According to the constant amplitude characteristic of the ZC sequence, the PRACH signal frame can be converted to the frequency domain, and valid data can be extracted from it to obtain the ZC sequence part. Then, the change in amplitude of the above-mentioned ZC sequence is mainly caused by the RF fingerprint and the channel effect. Since the RF fingerprint is stable and the channel effect is random, the influence of the channel can be separated and eliminated from the sampling, so that the influence of the RF fingerprint can be measured from the jitter of the ZC sequence. Based on the above ideas, the LTE PRACH RF fingerprint can be extracted according to the following steps.

[0044] Step 4: Extract the ZC sequence in the frequency domain. Perform the CP removal operation on the synchronized and frequency-debiased PRACH signal frame to obtain the leading sequence part, which is expressed as

[0045]

[0046] Where s(n) is the PRACH signal at the transmitter, h(n) and η(n) are the channel and noise, represents convolution. Performing Fast Fourier Transform (FFT) on r(n) yields

[0047] R(ω)=W N [FFT(r(n))] = W N [(S(ω)+σ'(ω))H(w)] (6)

[0048] in W N [g] represents a window of length N. Therefore, through FFT transformation, the convolution channel effect in equation (5) is transformed into a multiplication effect.

[0049] Step 5: Further transform the multiplicative channel effect into additive effect through cepstrum operation

[0050] C(ω)=log[R(ω)]=W N [log[S(ω)+σ'(ω)]+log[H(ω)]] (7)

[0051] Note that the channel fading H(ω) at different frequencies ω is different and random.

[0052] Step 6: Further perform cross-correlation operation on the above cepstrum

[0053]

[0054] in Indicates rounding down, N ZCIndicates the length of the frequency domain ZC sequence.

[0055] It should be noted that the above embodiments are not intended to limit the protection scope of the present invention, and equivalent changes or substitutions made on the basis of the above technical solutions all fall within the protection scope of the claims of the present invention.

Claims

1. A radio frequency fingerprint extraction method based on LTE PRACH signal cepstrum and cross-correlation operation, characterized in that: The method comprises the following steps: Step 1: Sample the LTE PRACH signal. Step 2: Synchronize the sampled signal. Step 3: Carrier frequency offset estimation and correction, Step 4: Perform fast Fourier transform on the frequency-debiased signal and extract N ZC Valid subcarrier data, Get the frequency domain ZC sequence, Step 5: Take the logarithm of the frequency domain ZC sequence obtained in step 4 to obtain the cepstrum domain signal. Step 6: Convert the previous Signal and after The cross-correlation operation is performed on the signals, and the result is used as the RF fingerprint. Among them, step 1: sample the LTE PRACH signal, denoted as r(n), Step 2: Based on the cyclic prefix structure of the PRACH signal frame, the following formula is used to achieve symbol synchronization in time: Where r(n) is the sampled PRACH signal, N CP is the sample length of the cyclic prefix, N SEQ is the sample length of the preamble sequence part in the PRACH frame, and N PRACH is the total sample length of the PRACH frame. The s value obtained according to formula (1) is the first sample point of the actual valid data. Step 3: Carrier frequency offset estimation and correction. Using the cyclic prefix structure of the PRACH signal frame, the phase deviation of the signal sampling is first estimated as Where k = 1L L and L is the number of repeated PRACH frames, then the carrier frequency offset can be calculated as where f s As the sampling rate, the signal sampling carrier frequency offset compensation can be Step 4: Extract the ZC sequence in the frequency domain. Perform the CP removal operation on the synchronized and frequency-debiased PRACH signal frame to obtain the leading sequence part, which is expressed as Where s(n) is the PRACH signal at the transmitter, h(n) and η(n) are the channel and noise, represents convolution, then, by performing Fast Fourier Transform (FFT) on r(n), we can get R(ω)=W N [FFT(r(n))]=W N [(S(ω)+σ'(ω))H(w)] (6) in W N [g] represents a window of length N. Therefore, through FFT transformation, the convolution channel effect in equation (5) is transformed into a multiplicative effect. Step 5: Further transform the multiplicative channel effect into an additive effect through cepstrum operation. C(ω)=log[R(ω)]=W N [log[S(ω)+σ'(ω)]+log[H(ω)]] (7) Note that the channel fading H(ω) at different frequencies ω is different and random, Step 6: Further perform cross-correlation operation on the above cepstrum in Indicates rounding down, N ZC Indicates the length of the frequency domain ZC sequence.

Images