Beidou signal deception detection system and method based on self-correlation time-frequency characteristics

Abstract

The invention relates to the technical field of satellite navigation signal processing and anti-interference, and discloses a Beidou signal deception detection system and method based on self-correlation time-frequency characteristics, and the method comprises the steps: receiving a Beidou signal, and carrying out the operation of the Beidou signal and a local pseudo-random code to generate a self-correlation function sequence; smoothing the sequence, and extracting time domain features such as a main-to-auxiliary peak ratio representing waveform structure distortion and frequency domain features such as spectrum flatness representing spectrum distribution difference; performing normalization and weighted fusion on the multi-dimensional features by using a fuzzy logic theory to obtain a comprehensive decision amount; in combination with a continuous consistency check, the decision amount is compared with a threshold to determine deception jamming. According to the invention, through time-frequency conjoint analysis and a fuzzy judgment mechanism, weak structure change caused by a deception signal is effectively captured, and the detection sensitivity and stability of the Beidou receiver for resisting small-delay and high-concealment deception attacks can be improved without additional hardware.

Description

Technical Field

[0001] This invention relates to the field of satellite navigation signal processing and anti-interference technology, specifically to a BeiDou signal deception detection system and method based on autocorrelation time-frequency characteristics. Background Technology

[0002] Currently, the BeiDou Navigation Satellite System has become a core support for critical infrastructure in transportation, power dispatching, and national defense. However, due to the relatively weak power of satellite signals when they reach the ground over long distances, and the relatively open nature of civilian signal systems, they are susceptible to malicious interference. In particular, generative spoofing jamming, by transmitting forged signals with structural parameters highly similar to the real signal, can covertly induce receivers to lock onto incorrect code phases or carrier frequencies. This type of attack often tamperes with spatiotemporal information without the user's knowledge, posing a serious challenge to the reliability of navigation and positioning.

[0003] To address the aforementioned security threats, existing defense strategies primarily focus on signal quality monitoring and parameter anomaly analysis. Conventional solutions typically utilize multiple correlators configured within the receiver to calculate the symmetry or flatness of the correlation function, thereby determining the presence of abnormal waveforms. Some solutions attempt to distinguish between genuine satellite signals from different directions and spoofed ground signals by monitoring for abrupt changes in the carrier-to-noise ratio of the received signal, or by using multi-antenna arrays to measure the consistency of the signal's angle of arrival, thus identifying the interference source during the baseband processing stage.

[0004] However, existing detection methods still have shortcomings in practical applications. Methods relying on single time-domain waveform features or power indices are prone to missed detections due to feature overwhelming when faced with highly similar spoofing signals with small delays and superimposed signals, because the distortion features are weak. Meanwhile, decision logic using fixed hard thresholds is difficult to adapt to dynamically changing electromagnetic noise environments, and random signal fluctuations often lead to an increased false alarm rate. Furthermore, while multi-antenna array-based schemes offer better detection performance, they are limited by high hardware costs, large size, and complex signal processing, making it difficult to achieve low-cost software deployment and upgrades on a large number of existing single-antenna receivers.

[0005] Therefore, this invention provides a BeiDou signal deception detection system and method based on autocorrelation time-frequency characteristics to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a BeiDou signal deception detection system and method based on autocorrelation time-frequency features. This solves the problems of insufficient sensitivity to detect small-delay, highly covert attacks, poor anti-interference robustness, and difficulty in low-cost deployment due to reliance on single-dimensional features in existing BeiDou signal deception detection technologies.

[0007] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for detecting BeiDou signal spoofing based on autocorrelation time-frequency characteristics, comprising the following steps performed in sequence: First, the BeiDou signal reception and autocorrelation function calculation steps are performed. The BeiDou analog signal is received and the discretized digital intermediate frequency signal is obtained. The local pseudo-random code sequence is used to perform sliding correlation operation with the digital intermediate frequency signal to generate an autocorrelation function sequence. Secondly, the time-domain feature extraction step of the autocorrelation function is performed, the amplitude of the autocorrelation function sequence is calculated and smoothed, and a set of time-domain features that can characterize waveform structure distortion is extracted. Next, the autocorrelation function frequency domain feature extraction step is performed to map the autocorrelation function sequence to the frequency domain and extract a set of frequency domain features that can characterize the differences in spectral distribution. Finally, the feature normalization and fuzzy fusion decision steps are performed. The time-domain feature set and the frequency-domain feature set are combined, and the combined features are fused and calculated using fuzzy logic theory to obtain a comprehensive decision value. The comprehensive decision value is then compared with a preset decision threshold: when the comprehensive decision value is greater than or equal to the decision threshold, it is determined that there is spoofing interference in the current received signal and an alarm signal is output; when the comprehensive decision value is less than the decision threshold, it is determined that the current received signal is a normal signal.

[0008] By adopting the above technical solution, the strategy of joint analysis of time domain and frequency domain features makes up for the deficiency that traditional single-domain detection cannot fully describe signal distortion; at the same time, by using fuzzy logic to fuse multi-dimensional features, the problem of false judgment in critical state of hard threshold decision is avoided. Therefore, a high-sensitivity identification capability of Beidou deception signal is obtained in complex electromagnetic environment, and the false alarm rate and missed alarm rate are reduced.

[0009] Preferably, in the BeiDou signal reception and autocorrelation function calculation step, the sliding correlation operation is specifically performed as follows: traversing the code phase offset within a preset code phase search range, performing a dot product between the discretized digital intermediate frequency signal and the local pseudo-random code sequence with different code phase offsets and accumulating the results, thereby calculating the correlation value corresponding to each code phase offset, and constructing the autocorrelation function sequence reflecting the distribution of signal energy on the time delay axis from all the correlation values.

[0010] By adopting the above technical solution, the fine distribution structure of signal energy in the code phase dimension can be accurately captured, providing a complete data foundation for subsequent feature extraction.

[0011] Preferably, in the time-domain feature extraction step of the autocorrelation function, the process of extracting the time-domain feature set includes: determining the maximum value and its corresponding time delay position in the smoothed correlation amplitude function through global extremum search, which are respectively used as the amplitude of the main correlation peak and the position of the main peak; counting the number of sampling points other than the main peak position that satisfy the local maximum condition and whose amplitude exceeds the relative threshold to obtain the number of secondary peaks; obtaining the largest amplitude value among all secondary peaks, and calculating the quotient of the amplitude of the main correlation peak and the largest amplitude value among all secondary peaks to obtain the amplitude ratio of the main and secondary peaks.

[0012] By adopting the above technical solution, it is possible to quantify the amplitude anomalies and multi-peak effects of relevant peaks, and effectively identify waveform splitting phenomena caused by multipath or deceptive signals.

[0013] Preferably, the time-domain feature set further includes peak broadening features, peak symmetry index, and main peak energy concentration index; the extraction process includes: searching for time points when the amplitude drops to half the peak amplitude on both sides of the main peak position, and calculating the time difference between the two time points as the peak broadening features; selecting equidistant sampling points on both sides of the main peak position, and calculating the ratio of the sum of the absolute values ​​of the amplitude differences at the symmetrical positions to the sum of the amplitudes to obtain the peak symmetry index; calculating the ratio of the signal energy in the neighborhood of the main peak position to the total signal energy in the entire search range to obtain the main peak energy concentration index.

[0014] By adopting the above technical solution, the geometric shape of the relevant peaks is finely characterized from three dimensions: peak width, symmetry and energy focusing. It can keenly capture the slight waveform asymmetry and broadening caused by the superposition of deceptive signals.

[0015] Preferably, in the step of extracting the frequency domain features of the autocorrelation function, the process of extracting the frequency domain feature set includes: mapping the autocorrelation function sequence to the frequency domain through discrete Fourier transform and performing energy normalization; counting the number of frequency points in the normalized spectrum distribution that satisfy the local maximum condition and whose amplitude exceeds the relative threshold in the frequency domain, to obtain the number of spectral peaks; calculating the frequency corresponding to the energy centroid of the normalized spectrum distribution using the centroid method, to obtain the center frequency position; and calculating the second-order central moment of the normalized spectrum distribution around the center frequency position, to obtain the spectral spread feature.

[0016] By adopting the above technical solution, it is possible to detect abnormal frequency components, center frequency shift, and spectral energy dispersion of signals in the frequency domain, and identify deceptive interference with specific frequency domain fingerprint characteristics.

[0017] Preferably, the frequency domain feature set further includes an effective bandwidth range, a spectral flatness index, and a spectral symmetry index; the extraction process includes: searching for the smallest continuous frequency band interval where the proportion of accumulated energy to total energy reaches a preset energy coverage ratio, calculating the width of the smallest continuous frequency band interval as the effective bandwidth range; calculating the ratio of the geometric mean to the arithmetic mean of the normalized spectral distribution amplitude to obtain the spectral flatness index; selecting equidistant frequency points on the left and right sides of the center frequency position, calculating the degree of difference in spectral amplitude at symmetrical positions to obtain the spectral symmetry index.

[0018] By adopting the above technical solution, noise interference and structured spoofing signals can be further distinguished by the flatness and bandwidth characteristics of the spectrum, thereby enhancing the detection algorithm's ability to distinguish different types of interference.

[0019] Preferably, in the feature normalization and fuzzy fusion decision step, the process of calculating the comprehensive decision value using fuzzy logic theory includes: firstly, using preset statistical extreme value parameters to perform numerical normalization on the features; for features with larger values ​​that are more likely to be abnormal, the difference between the current feature value and the historical statistical minimum value is divided by the difference between the historical statistical maximum value and the historical statistical minimum value for calculation; for features with smaller values ​​that are more likely to be abnormal, the difference between the historical statistical maximum value and the current feature value is divided by the difference between the historical statistical maximum value and the historical statistical minimum value for calculation; then, using preset trapezoidal membership function or S-shaped membership function, the normalized feature values ​​are mapped to the numerical range of zero to one to obtain the deception state membership value; finally, the deception state membership values ​​of all features are weighted and summed according to preset weight coefficients.

[0020] By adopting the above technical solution, features with different physical dimensions are uniformly mapped into a probability measure of deception possibility, and the confidence of the judgment result is improved through weighted fusion.

[0021] Preferably, the method further includes continuous consistency detection logic: after comparing the comprehensive decision quantity with the decision threshold, it is determined whether the comprehensive decision quantity of multiple consecutive processing cycles is greater than or equal to the decision threshold, and the alarm signal is confirmed to be output only when the condition is met.

[0022] By adopting the above technical solution, the continuous verification in the time dimension filters out accidental misjudgments caused by transient noise, thereby improving the anti-interference stability of the system.

[0023] Secondly, the present invention provides a BeiDou signal spoofing detection system based on autocorrelation time-frequency characteristics, comprising: The BeiDou signal receiving and processing module is used to receive BeiDou analog signals and obtain discrete digital intermediate frequency signals. It uses an internally preset local pseudo-random code sequence to perform sliding correlation operation with the digital intermediate frequency signals and outputs an autocorrelation function sequence. The time-domain feature extraction module is used to perform amplitude calculation and smoothing on the autocorrelation function sequence, and extract a set of time-domain features that can characterize waveform structure distortion. The frequency domain feature extraction module is used to map the autocorrelation function sequence to the frequency domain and extract a set of frequency domain features that can characterize the differences in spectral distribution. The fuzzy fusion decision module is used to combine the time-domain feature set and the frequency-domain feature set, use fuzzy logic theory to perform fusion calculation on the combined features to obtain a comprehensive decision quantity, and compare the comprehensive decision quantity with a preset decision threshold to determine whether there is deception interference in the currently received signal.

[0024] By adopting the above technical solutions, a system architecture integrating signal preprocessing, multi-domain feature analysis and intelligent fusion decision was constructed, realizing real-time and accurate detection of BeiDou navigation signal deception interference.

[0025] This invention provides a BeiDou signal spoofing detection system and method based on autocorrelation time-frequency characteristics. It has the following beneficial effects: 1. This invention constructs a multi-dimensional signal quality fingerprint by jointly extracting the time-domain feature set and frequency-domain feature set of the autocorrelation function. This time-frequency joint analysis mechanism overcomes the limitations of single-domain detection, and can keenly capture the subtle waveform distortion and spectral distribution differences caused by the superposition of deceptive signals and real signals, thereby improving the detection sensitivity and recognition accuracy against covert deception attacks with small delays and high similarity.

[0026] 2. This invention introduces a fusion decision mechanism based on statistical extreme value normalization and fuzzy logic theory, mapping feature values ​​with different physical dimensions to deceptive state membership degrees for weighted summation. This mechanism replaces the traditional hard threshold decision method, effectively solving the problem of decision jumps easily occurring under critical conditions with a single feature, and reducing the risk of false alarms caused by environmental noise through the complementarity of multiple features, thereby enhancing the robustness and stability of the detection system operating in complex electromagnetic environments.

[0027] 3. The detection logic employed in this invention is entirely based on the discrete digital intermediate frequency signal and its autocorrelation function sequence generated internally by the receiver. It requires no modification to the physical layer architecture of the BeiDou navigation signal and does not rely on additional hardware support such as multi-antenna arrays. This method can be integrated into the existing BeiDou receiver baseband processing flow as a pure software algorithm, offering advantages such as low computational complexity, no occupation of radio frequency hardware resources, low engineering implementation cost, and ease of upgrading and deploying existing equipment. Attached Figure Description

[0028] Figure 1 This is a system architecture diagram of an embodiment of the present invention; Figure 2 This is a flowchart of a method according to an embodiment of the present invention; Figure 3 This is a schematic diagram showing the comparison of the success rate of BeiDou signal deception detection under different typical deception scenarios according to an embodiment of the present invention; Figure 4 This diagram illustrates a comparison of the detection performance of the method of this invention and the traditional correlation function distortion detection method under the same false alarm probability conditions. Detailed Implementation

[0029] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] See attached document Figure 1 This invention provides a BeiDou signal spoofing detection system based on autocorrelation time-frequency characteristics, which is deployed in the signal processing link of a BeiDou satellite navigation receiver. This system can be integrated into a digital signal processor, field-programmable gate array (FPGA), or general-purpose processor through software algorithms or firmware logic without altering the existing BeiDou signal architecture. The system mainly includes a BeiDou signal receiving and processing module, a time-domain feature extraction module, a frequency-domain feature extraction module, and a fuzzy fusion decision module.

[0031] The BeiDou signal receiving and processing module, serving as the system's front-end input, connects to the receiver's RF front-end or analog-to-digital converter (ADC). This module receives the BeiDou analog signal after antenna acquisition and RF front-end processing, performs down-conversion to remove carrier frequency components, and then performs analog-to-digital conversion sampling to output a discrete digital intermediate frequency (IF) signal sequence. The BeiDou signal receiving and processing module internally includes a local pseudo-random code generator to generate a local pseudo-random code sequence corresponding to the currently detected satellite channel. This module performs a sliding correlation operation on the received discrete digital signal and the local pseudo-random code sequence, calculating the correlation value under different code phase offsets, thereby outputting the autocorrelation function sequence of the BeiDou signal. This autocorrelation function sequence reflects the temporal alignment between the received signal and the local code, as well as the signal's waveform characteristics.

[0032] The time-domain feature extraction module is connected to the BeiDou signal receiving and processing module to receive autocorrelation function sequences and perform time-domain analysis. This module first calculates the amplitude of the autocorrelation function, obtaining the correlation amplitude function. To reduce the interference of environmental noise on the accuracy of feature extraction, the time-domain feature extraction module performs smoothing filtering on the correlation amplitude function. Based on the smoothed autocorrelation amplitude curve, the module extracts multi-dimensional time-domain features through extremum search and waveform analysis algorithms. These time-domain features specifically include the amplitude information of the main correlation peak, the number of secondary peaks besides the main peak, the amplitude ratio between the main correlation peak and the largest secondary peak, the peak broadening characteristics of the main correlation peak at a specific height, the symmetry index of the waveforms to the left and right of the main correlation peak, and the concentration index of the energy in the main correlation peak region relative to the overall energy. These features quantify the degree of waveform distortion of the autocorrelation function in the time domain from different perspectives.

[0033] The frequency domain feature extraction module is also connected to the output of the time domain feature extraction module or independently connected to the smoothed autocorrelation amplitude data stream. This module is used to convert the time-domain autocorrelation amplitude function into a frequency domain expression to mine the spectral features of the signal. Specifically, this module performs a Discrete Fourier Transform on the smoothed autocorrelation amplitude sequence to obtain the spectral sequence of the autocorrelation function, and performs energy normalization processing on the spectral sequence to eliminate the influence of total signal power fluctuations. Based on the normalized spectrum, the frequency domain feature extraction module calculates and extracts multi-dimensional frequency domain features. These frequency domain features specifically include the number of significant peaks in the spectrum, the center frequency position of the spectral energy distribution, the spread feature reflecting the degree of spectral dispersion, the effective bandwidth range containing the main energy, the flatness index measuring the flatness of the spectrum, and the symmetry index of the spectral distribution. These features are used to capture frequency domain structural anomalies caused by the superposition of deceptive signals.

[0034] The fuzzy fusion decision module is connected to the time-domain feature extraction module and the frequency-domain feature extraction module, serving as the system's decision output unit. This module receives the time-domain and frequency-domain feature sets from the two modules mentioned above and combines them to form a high-dimensional feature vector. The fuzzy fusion decision module first normalizes each feature dimension using preset statistical extreme value parameters, mapping physical quantities of different dimensions to a unified numerical range. Then, based on fuzzy logic theory, the module uses a preset membership function to map the normalized feature values ​​to membership values ​​specific to the deception state. This membership value characterizes the degree to which the current feature tends to indicate a deception signal. The fuzzy fusion decision module assigns corresponding weight coefficients based on the contribution of each feature to deception detection and performs a weighted summation of the deception membership degrees of all features to obtain a comprehensive decision quantity. This module compares the comprehensive decision quantity with a preset decision threshold. When the comprehensive decision quantity exceeds the threshold, it determines that deception interference exists in the currently received signal and outputs an alarm signal; otherwise, it determines that the signal is normal. In addition, this module can be configured with continuous consistency detection logic, which requires that the comprehensive decision amount of multiple consecutive processing cycles exceeds the threshold before confirming the output of a deceptive alarm, so as to reduce the false alarm rate of the system.

[0035] See attached document Figure 2 This invention provides a BeiDou signal spoofing detection method based on autocorrelation time-frequency features. This method identifies spoofing attacks by jointly analyzing the feature differences of the signal in the time and frequency domains and using fuzzy logic for comprehensive decision-making. The method mainly includes the following four sequentially executed processing steps: BeiDou signal reception and autocorrelation function calculation, time-domain feature extraction of the autocorrelation function, frequency-domain feature extraction of the autocorrelation function, and feature normalization and fuzzy fusion decision-making.

[0036] In the BeiDou signal reception and autocorrelation function calculation steps, the receiver first performs down-conversion and analog-to-digital conversion on the BeiDou analog signal captured by the RF front-end to obtain a discretized digital intermediate frequency (IF) signal. Then, a sliding correlation operation is performed between the receiver-generated pseudo-random code sequence and this IF signal. This operation traverses a set code phase search range, calculating the correlation value corresponding to each code phase offset point, thereby generating an autocorrelation function sequence reflecting the distribution of signal energy along the time delay axis. This autocorrelation function sequence serves as the fundamental data source for subsequent feature extraction and analysis.

[0037] In the time-domain feature extraction step of the autocorrelation function, the system calculates its amplitude function based on the autocorrelation function generated in the previous step and performs smoothing filtering on the amplitude function to suppress noise interference. Based on the smoothed time-domain waveform, the system uses waveform analysis algorithms to extract multi-dimensional time-domain features that can characterize the distortion of the correlation peak structure. The time-domain feature set specifically includes the amplitude of the main correlation peak, the number of secondary peaks existing outside the neighborhood of the main peak, the amplitude ratio between the main correlation peak and the largest secondary peak, the waveform width at the half-peak of the main correlation peak, the symmetry index on the left and right sides of the main correlation peak, and the concentration index of the energy of the main peak region relative to the energy of the entire search range.

[0038] In the autocorrelation function frequency domain feature extraction step, the system maps the smoothed time-domain autocorrelation amplitude function to the frequency domain through a discrete Fourier transform, and performs energy normalization on the generated spectral sequence. On the normalized spectral distribution, the system extracts multidimensional frequency domain features that characterize differences in spectral shape. Specifically, the set of frequency domain features includes the number of significant peaks in the spectrum, the center frequency position of the spectral energy distribution, the spread reflecting the degree of spectral dispersion, the effective bandwidth range containing the main signal energy, the flatness index measuring the degree of spectral amplitude fluctuation, and the symmetry index of the spectral distribution about the center frequency.

[0039] In the feature normalization and fuzzy fusion decision steps, the system first combines the extracted time-domain feature set and frequency-domain feature set, and uses statistical extreme values ​​to normalize each feature dimension, mapping all feature values ​​to a unified numerical range. Then, based on fuzzy logic theory, the system calculates the membership degree value of each normalized feature belonging to the deception state using a preset membership function. The system assigns corresponding weight coefficients according to the contribution of each feature to deception detection, and performs a weighted summation of the deception membership degrees of all features to obtain a comprehensive decision value. Finally, the system compares this comprehensive decision value with a preset decision threshold. When the comprehensive decision value is greater than or equal to the preset threshold, it determines that deception interference exists in the currently received signal and outputs an alarm message; otherwise, it determines that the signal is a genuine and normal signal.

[0040] In the BeiDou signal reception and autocorrelation function calculation steps, the first step is to acquire the digital baseband or intermediate frequency (IF) signal to be processed. The BeiDou receiver captures the weak analog BeiDou signal in space via an RF antenna. This analog signal undergoes low-noise amplification, bandpass filtering, and down-conversion processing by the RF front-end circuit, shifting the carrier frequency to the IF band. Subsequently, an analog-to-digital converter (ADC) is used to discretize the analog IF signal, converting the continuous-time analog signal into a discrete signal sequence in the digital domain. In this embodiment of the invention, the discrete signal obtained after the above down-conversion and sampling processing is denoted as... .

[0041] Simultaneously, the receiver generates a corresponding local pseudo-random code sequence based on the currently tracked or acquired satellite number using a local signal generator. This local pseudo-random code sequence is a standard reference signal pre-stored or generated in real-time by the receiver, used for matched filtering with the received mixed signal. This local pseudo-random code sequence is denoted as... To assess the similarity and time alignment between the received signal and the local reference signal, the system performs relevant computational processing. Specifically, the system processes the received discrete signal... Perform dot products with local pseudo-random code sequences of different delay phases and accumulate them.

[0042] The correlation operation is performed within a set coherent integration period, aiming to improve the signal-to-noise ratio by accumulating signal energy, thereby extracting useful signal features from the noisy background. The autocorrelation function obtained after the correlation operation... The mathematical expression is: ; in, The code phase delay index represents the time delay of the local pseudo-random code sequence relative to the received signal. The search window is varied to cover the possible signal arrival time range; This represents the length of the correlation integral, which is the total number of discrete sampling points involved in a single correlation operation. This depends on the system's sampling rate and the set coherent integration time. It involves traversing the search range. The system calculates a series of values ​​for each value. These values ​​constitute a complete autocorrelation function curve, which reflects the signal energy distribution under different code phase delays and serves as the foundational data for subsequent time-domain and frequency-domain feature analysis.

[0043] In the time-domain feature extraction step of the autocorrelation function, the system first performs modulus calculation on the complex autocorrelation function calculated in step one to obtain a real number sequence containing only amplitude information. Let the correlation amplitude function be... Since thermal noise and environmental interference are inevitably mixed into the received signal, directly using the original amplitude function for feature extraction may introduce errors. Therefore, the system uses a smoothing filter to process the autocorrelation amplitude function to smooth out glitches and preserve the main structure of the waveform. The smoothed autocorrelation amplitude function is shown below. The calculation formula is: ; in, These represent the coefficients of the smoothing filter, which determine the frequency response characteristics of the filter. Gaussian window or Hanning window coefficients are usually selected. This represents the half-width of the smoothing window, used to control the range of smoothing.

[0044] Based on the smoothed autocorrelation amplitude function The system first performs an extremum search to determine the characteristics of the main correlation peak. The main correlation peak typically corresponds to the energy concentration point of the real signal or the strongest deceptive signal. The system iterates through... The value of is taken, and its global maximum value is extracted as the amplitude of the main correlation peak. And record the corresponding time delay position. The relevant mathematical expression is as follows: ; ; Subsequently, the system detects secondary peaks in addition to the main correlation peak. The presence of secondary peaks often indicates the superposition of multipath effects or deceptive signals. The system defines points that satisfy the local maximum condition and whose amplitude exceeds a certain threshold as valid peak points. Specifically, for any sampling point... If it meets the following conditions, it is determined to be a peak point: ; In the above-mentioned logic of judgment, This indicates the width of the local neighborhood, used to exclude small fluctuations near the same peak. This represents the relative threshold coefficient, ranging from 0 to 1, used to filter out spurious peaks caused by background noise. The system will exclude the main peak... All peaks other than those meeting the above conditions constitute the set of secondary peaks. The number of secondary peaks is then counted to obtain the characteristic of the number of secondary peaks. : ; To quantify the relative intensity relationship between the main peak and the strongest interfering component, the system calculates the amplitude ratio of the main and secondary peaks. The system first performs calculations on the secondary peak set... Find the secondary peak with the largest amplitude, and denot its amplitude as . The calculation formula is: ; Based on this, the amplitude ratio of the main peak and the secondary peak is defined. It is the quotient of the main peak amplitude and the largest secondary peak amplitude. To prevent the denominator from being zero due to the absence of a secondary peak, a very small positive number is introduced. The correction is made, and the calculation formula is as follows: ; The superposition of deceptive signals not only generates secondary peaks but may also distort the shape of related peaks, such as broadening or asymmetry. Therefore, the system further extracts peak broadening features. First, the half-peak amplitude is defined. Half the amplitude of the main correlation peak, i.e. Then at the main peak position The time points at which the search amplitudes on both the left and right sides decrease to the half-peak amplitude are respectively recorded as the left intersection point. and right intersection point satisfy Peak broadening characteristics Defined as the width between these two intersection points: ; To detect asymmetric distortion of waveforms, the system constructs a peak shape symmetry index. This indicator quantifies symmetry by calculating the difference in amplitude values ​​at equidistant locations on either side of the main peak. Specifically, the calculation is performed at the main peak location... Take from both sides The calculation formula is as follows: (Number of sampling points) ; In the formula, The absolute values ​​of the amplitude differences between the left and right symmetrical points are summed. The sum of the amplitudes of the left and right symmetrical points is used for normalization. To prevent extremely small positive numbers with a denominator of zero, this index approaches zero when the waveform is perfectly symmetrical; its value increases when asymmetric distortion exists. Furthermore, the system also calculates the main peak energy concentration characteristics. This characteristic reflects the degree of signal energy focusing near the main peak. It is defined as the neighborhood of the main peak. The ratio of the signal energy within the search area to the total signal energy across the entire search range, where Let be the radius of the defined energy integration interval. The calculation formula is: ; Finally, the system combines the extracted feature parameters in a predetermined order to form a time-domain feature vector describing the current signal state. This vector includes components such as the amplitude of the main correlation peak, the number of secondary peaks, the ratio of main to secondary peak amplitudes, peak broadening characteristics, peak symmetry index, and the energy concentration of the main peak. Additional features such as peak spacing or peak offset can be added according to actual engineering requirements. The expression for the feature vector is: ; In the frequency domain feature extraction step of the autocorrelation function, the system extracts the smoothed correlation amplitude function obtained in the previous step. A time-frequency domain transformation is performed to extract the signal's distribution characteristics in the frequency domain. The frequency domain feature extraction module uses Discrete Fourier Transform (DFT) or Fast Fourier Transform (FFT) algorithms to process the time-domain sequence, mapping the time-domain energy distribution to the frequency-domain spectral density. The calculated frequency domain representation... As shown below: ; in, For code phase delay index; The number of code phase search points for the autocorrelation function; This is a frequency index. To eliminate the influence of total power fluctuations in the received signal on the spectral shape characteristics and ensure the comparability of characteristics under different signal strengths, the system indexes the spectrum. Energy normalization is performed. The normalized spectrum is shown below. The calculation formula is: ; in, The total energy of the spectrum. To prevent extremely small positive numbers with a denominator of zero, based on this normalized spectrum... The system extracts a series of multidimensional features that can characterize the spectral distortion caused by deception interference.

[0045] First, the system extracts the peak count feature of the spectrum. Similar to the time-domain peak extraction logic, the system counts the number of frequency points in the spectrum that satisfy the local maximum condition and whose amplitude exceeds a certain relative threshold. When multipath or spoofing signals are present, the spectrum often exhibits multi-peak characteristics. (The last sentence, "Number of Spectral Peaks," appears to be an error and is left untranslated.) The calculation formula is: ; in, This is the relative threshold coefficient for frequency domain peak decision, used to remove spurious peaks in the spectral noise floor.

[0046] Secondly, the system calculates the center frequency characteristic of the spectrum, which reflects the location of the signal's energy centroid. For normal signals, the energy centroid is usually located near the carrier center frequency; however, the introduction of deceptive signals may cause a shift in the energy centroid. Center Frequency Calculation using the centroid method: ; The system further extracts the spectral spread feature, which uses the second-order central moment of the spectrum to measure the dispersion of spectral energy around the center frequency. This indicator increases significantly when the spoof signal is superimposed on the real signal, causing spectral broadening. Spectral Spread The calculation formula is: ; To assess the bandwidth occupancy of signal energy, the system extracts effective bandwidth features. The system searches the spectrum for a minimum continuous frequency band. This ensures that the cumulative energy within this interval reaches the preset energy coverage ratio. The criterion for satisfying the condition is: ; Based on this, effective bandwidth Defined as the width of this frequency band: ; The system also calculates spectral flatness features, used to describe the flatness of the spectral distribution. Spectral flatness is defined as the ratio of the geometric mean to the arithmetic mean of the spectral amplitudes. Superposition of deceptive signals often alters the spectral structure of the signal, changing its flatness. Spectral flatness The calculation formula is as follows: ; It should be noted that in the above formula... It refers to the total number of frequency points involved in the calculation, to distinguish it from the previously defined characteristic of the number of spectral peaks.

[0047] Furthermore, the system extracts a spectral symmetry index to quantify the degree of left-right symmetry of the spectrum about the center frequency. The system at the center frequency... Take from both the left and right sides For each frequency point, calculate the difference in spectral amplitude at the corresponding location. Spectral symmetry index. The calculation formula is: ; Finally, the system combines the extracted spectral peak count, center frequency, spectral spread, effective bandwidth, spectral flatness, and spectral symmetry indices to construct a frequency domain feature set. This feature set comprehensively describes the signal's distribution characteristics in the frequency domain, and its expression is: ; In the feature normalization and fuzzy fusion decision step, the system processes the temporal feature vectors extracted in the previous step. and frequency domain eigenvectors The data is then integrated and analyzed to generate the final deception detection result. First, the two feature vectors are merged to form a single vector containing... The total eigenvector of the dimensional features : ; Because the physical meanings and numerical ranges of the various features differ, there are dimensional differences. To eliminate this difference and improve the universality of subsequent membership functions and decision thresholds, the system modifies the total feature vector. Each dimension of features Normalization is performed. In this embodiment, the min-max normalization method is used to map the original feature values ​​to the [0,1] interval. For features where larger values ​​indicate a higher likelihood of anomalies, such as the number of secondary peaks... Or peak broadening Its normalization characteristics The calculation formula is: ; in, and The first The historical statistical minimum and maximum values ​​of the dimensional features under normal signal conditions can be obtained through offline training or online statistics. To prevent extremely small positive numbers with a denominator of zero, the smaller the value, the more likely it is to be an anomaly, such as the ratio of the amplitude of the main peak to the sub-peak. Then, reverse normalization is used, and its calculation formula is: ; After normalization, all features are standardized to the same scale, and their numerical values ​​directly reflect the degree to which they deviate from the normal state. Next, based on fuzzy logic theory, the system normalizes each dimension of the features. Construct a membership function for the deception state. This function converts the precise feature value into a membership degree in the interval [0,1], representing the degree to which the feature indicates the likelihood of a deception attack. In this embodiment, a trapezoidal membership function can be used, and its mathematical expression is: ; in, and The deception decision inflection point parameter for this feature defines the range in which the feature value transitions from a state that is not at all deceptive to a state that is fully deceptive. Correspondingly, the membership degree of the normal state can be defined as... To achieve a smoother transition, an S-shaped membership function can also be used: ; in, The slope of the control function, i.e., the steepness of the transition; This represents the inflection point of the function. After obtaining the membership degrees of all features to the deception state, the system weights and fuses these membership values ​​to calculate a comprehensive decision value. This comprehensive judgment score fully reflects the degree of deception risk indicated by all characteristics. Its calculation formula is as follows: ; in, For the first The weights of each feature are non-negative, and the sum of all weights satisfies Weight The value can be determined based on the contribution of this feature in distinguishing normal samples from deceptive samples, or it can be preset through offline calibration or expert knowledge.

[0048] Finally, the system will calculate the comprehensive decision value. With a preset decision threshold Compare. If If the signal is detected by a spoofing attack, the system determines that the currently received signal is normal; otherwise, the signal is considered normal. To improve the stability and reliability of the decision and avoid false alarms caused by transient interference, the system can also introduce a continuity check rule. This rule requires that in continuous... Within each observation window, the calculated comprehensive decision value all satisfy... Only then will a deceptive alarm signal be finally output. This mechanism effectively filters out the influence of a single or a few anomalies, enhancing the robustness of the detection method.

[0049] To verify the effectiveness and reliability of the proposed BeiDou signal spoofing detection system and method based on autocorrelation time-frequency features in practical application scenarios, this embodiment constructs a composite signal test environment containing both real signals and spoofing signals at the same frequency. The BeiDou signal sampling rate used in the experiment was set to 25 MSps, and the coherence integration time was set to 1 ms. The received signal simulated a mixed input of real satellite signals and spoofing interference signals. The parameters of the spoofing signal included power, code delay, and carrier phase parameters. The system dynamically configured itself according to different test scenarios to generate corresponding spoofing attack samples. In the feature extraction stage, as a preferred embodiment of this invention, in addition to extracting the aforementioned basic time-frequency features, a four-level discrete wavelet transform was performed on the autocorrelation function using the Symlets4 wavelet basis function, and the moving average statistics of the coefficients at each level were calculated as supplementary features. These statistics specifically include the mean and variance. These supplementary features, along with the basic features, are input to the fuzzy fusion decision module to enhance the system's ability to capture weak signal distortions.

[0050] See attached document Figure 3 This paper presents a comparison of the success rate of BeiDou signal deception detection under different typical deception scenarios according to an embodiment of the present invention. The experiment sets the false alarm probability to 10 to the power of -2 and tests are conducted on six typical deception attack scenarios. Figure 3 The horizontal axis represents different types of deception scenarios, and the vertical axis represents the percentage of successful detection. For example... Figure 3As shown, for the four scenarios of high-power time-progression spoofing, equal-power time-progression spoofing, frequency-locked position-progression spoofing, and time-progression spoofing under dynamic platforms, the height of the corresponding detection success rate histograms is close to or reaches the 100-point scale. This indicates that the method of the present invention has extremely high recognition accuracy when dealing with general spoofing attacks with significant power, significant time offset, or dynamic platform changes, and can achieve near-complete effective detection. For the more concealed carrier phase alignment spoofing scenario, because the spoofing signal and the real signal maintain a high degree of consistency in carrier phase, the distortion of the correlation peak waveform after superposition is extremely small, significantly increasing the detection difficulty. Figure 3 The bar chart showing the detection success rate in this scenario has a height of approximately 82 numerical increments. While this is lower than in typical scenarios, it remains within the effective detection range. For advanced zero-latency replay spoofing scenarios... Figure 3 The bar chart showing the detection success rate has a height of approximately 89 numerical increments. The above data indicates that this invention, through multi-dimensional feature fusion, still possesses good detection capabilities when dealing with complex deception attacks characterized by high similarity and high concealment.

[0051] See attached document Figure 4 The figure illustrates the receiver operating characteristic curves of the method of this invention and conventional detection methods under different false alarm probabilities. The horizontal axis represents the false alarm probability, using a logarithmic coordinate system, ranging from 10⁻⁴ to 10⁻¹; the vertical axis represents the detection probability, ranging from 0 to 0.9. The solid line represents the method of this invention, while the dashed and dotted lines represent two conventional detection methods relying on a single feature, respectively. Figure 4 As can be intuitively observed, throughout the entire false alarm probability range of the test, the solid curve corresponding to the method of this invention is always above the dashed and dotted curves corresponding to the traditional method. This means that under the same false alarm probability constraints, the method of this invention can achieve a higher detection probability than the traditional method. Especially under the strict condition of low false alarm probability, i.e., 10 to the power of -4 to 10 to the power of -3, the detection probability of the traditional method is lower, and the curve position is lower; while the method of this invention still shows a significant performance advantage in this range, and the curve position is significantly higher than that of the traditional method. This comparative result fully confirms that the present invention, by fusing multi-dimensional features in the time domain and frequency domain and introducing a fuzzy decision mechanism, effectively overcomes the deficiency of insufficient sensitivity of traditional single-feature methods in complex environments, and improves the robustness and overall detection performance of the system.

Claims

1. A method for detecting BeiDou signal spoofing based on autocorrelation time-frequency characteristics, characterized in that, Includes the following steps: S1. Steps for receiving BeiDou signals and calculating autocorrelation function: Receive BeiDou analog signals and obtain discretized digital intermediate frequency signals. Perform sliding correlation operation between the local pseudo-random code sequence and the digital intermediate frequency signals to generate an autocorrelation function sequence. S2. Autocorrelation function time-domain feature extraction step: Perform amplitude calculation and smoothing on the autocorrelation function sequence, and extract a set of time-domain features that can characterize waveform structure distortion; S3. Autocorrelation function frequency domain feature extraction step: Map the autocorrelation function sequence to the frequency domain and extract a set of frequency domain features that can characterize the differences in spectral distribution; S4. Feature normalization and fuzzy fusion decision step: Combine the time-domain feature set and the frequency-domain feature set, use fuzzy logic theory to perform fusion calculation on the combined features to obtain a comprehensive decision quantity, and compare the comprehensive decision quantity with a preset decision threshold: when the comprehensive decision quantity is greater than or equal to the decision threshold, it is determined that there is deception interference in the current received signal and an alarm signal is output. When the overall decision value is less than the decision threshold, the currently received signal is determined to be a normal signal.

2. The BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics according to claim 1, characterized in that, In step S1, the process of receiving the BeiDou analog signal and obtaining the discrete digital intermediate frequency signal includes: receiving the BeiDou analog signal after radio frequency front-end processing, performing down-conversion processing on the BeiDou analog signal to remove the carrier frequency component, and then performing analog-to-digital conversion sampling. The sliding correlation operation includes: traversing the code phase offset within a preset code phase search range, performing a dot product between the discretized digital intermediate frequency signal and the local pseudo-random code sequence with different code phase offsets and accumulating the results, and constructing the autocorrelation function sequence from all correlation values. The autocorrelation function sequence reflects the distribution of signal energy on the time delay axis.

3. The BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics according to claim 1, characterized in that, In step S2, the process of calculating the magnitude and smoothing the autocorrelation function sequence includes: The correlation amplitude function is obtained by performing a modulus operation on the autocorrelation function sequence in complex form. The correlation amplitude function is then weighted and summed within a set smoothing window using preset smoothing filter coefficients to obtain the smoothed correlation amplitude function.

4. The BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics according to claim 3, characterized in that, The time-domain feature set includes the amplitude of the main correlation peak, the number of secondary peaks, and the ratio of the amplitudes of the main and secondary peaks; the process of extracting the time-domain feature set includes: The maximum value and the corresponding time delay position in the smoothed correlation amplitude function are determined by global extremum search, and are respectively used as the amplitude of the main correlation peak and the position of the main peak. The number of secondary peaks is obtained by counting the number of sampling points other than the main peak location that satisfy the local maximum condition and whose amplitude exceeds the relative threshold. Obtain the largest amplitude value among all secondary peaks, calculate the quotient of the amplitude value of the main correlation peak and the largest amplitude value among all secondary peaks, and obtain the amplitude ratio of the main and secondary peaks.

5. The BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics according to claim 4, characterized in that, The time-domain feature set also includes peak broadening features, peak symmetry indices, and main peak energy concentration indices; the process of extracting the time-domain feature set further includes: At the time points when the search amplitude on both sides of the main peak position decreases to half the peak amplitude, the time difference between the two time points is calculated as the peak broadening feature. Equal sampling points are selected on the left and right sides of the main peak position, and the ratio of the sum of the absolute values ​​of the amplitude differences at the symmetrical positions to the sum of the amplitudes is calculated to obtain the peak shape symmetry index. The ratio of the signal energy in the neighborhood of the main peak location to the total signal energy within the entire search range is calculated to obtain the main peak energy concentration index.

6. The BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics according to claim 1, characterized in that, The frequency domain feature set includes the number of spectral peaks, the location of the center frequency, and the spectral spread characteristics. In step S3, the process of mapping the autocorrelation function sequence to the frequency domain includes: generating a spectral sequence using discrete Fourier transform, calculating the total energy of the spectral sequence, and dividing the amplitude value of each frequency point in the spectral sequence by the total energy to obtain a normalized spectral distribution; The process of extracting the frequency domain feature set includes: The number of frequency points in the normalized spectrum distribution that satisfy the local maximum condition and whose amplitude exceeds the relative threshold in the frequency domain is counted to obtain the number of spectrum peaks. The frequency corresponding to the energy centroid of the normalized spectral distribution is calculated using the centroid method to obtain the position of the center frequency. The second-order central moment of the normalized spectral distribution around the center frequency position is calculated to obtain the spectral spread characteristic.

7. The BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics according to claim 6, characterized in that, The frequency domain feature set also includes effective bandwidth range, spectral flatness index, and spectral symmetry index; the process of extracting the frequency domain feature set further includes: The search term is defined as the minimum continuous frequency band interval where the proportion of accumulated energy to total energy reaches a preset energy coverage ratio, and the width of the minimum continuous frequency band interval is calculated as the effective bandwidth range. The ratio of the geometric mean to the arithmetic mean of the normalized spectral distribution amplitude is calculated to obtain the spectral flatness index; Frequency points at equal intervals are selected on both sides of the center frequency position, and the degree of difference in spectral amplitude at the symmetrical positions is calculated to obtain the spectral symmetry index.

8. The BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics according to claim 1, characterized in that, In step S4, the process of using fuzzy logic theory to fuse and calculate the combined features to obtain the comprehensive decision value includes: The combined features are numerically normalized using preset statistical extreme value parameters: for features that are more likely to be abnormal, the current feature value is subtracted from the historical minimum value and then divided by the difference between the historical maximum value and the historical minimum value; for features that are more likely to be abnormal, the historical maximum value is subtracted from the current feature value and then divided by the difference between the historical maximum value and the historical minimum value. Using a preset trapezoidal or S-shaped membership function, the eigenvalues ​​after the numerical normalization process are mapped to a numerical range of zero to one to obtain the membership value of the deception state. The deception state membership values ​​of all features are weighted and summed according to preset weight coefficients to obtain the comprehensive decision value.

9. The BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics according to claim 1, characterized in that, In step S4, the method further includes continuous consistency detection logic: after comparing the comprehensive decision quantity with the preset decision threshold, it is determined whether the comprehensive decision quantity of multiple consecutive processing cycles is greater than or equal to the decision threshold; if so, the alarm signal is confirmed to be output; if not, the alarm signal is not output.

10. A BeiDou signal spoofing detection system based on autocorrelation time-frequency characteristics, characterized in that, Implementing the BeiDou signal spoofing detection method based on autocorrelation time-frequency characteristics as described in any one of claims 1-9, comprising: The BeiDou signal receiving and processing module is used to receive BeiDou analog signals and obtain discrete digital intermediate frequency signals. It uses an internally preset local pseudo-random code sequence to perform sliding correlation operation with the digital intermediate frequency signals and outputs an autocorrelation function sequence. The time-domain feature extraction module is used to perform amplitude calculation and smoothing on the autocorrelation function sequence, and extract a set of time-domain features that can characterize waveform structure distortion; the set of time-domain features includes the amplitude of the main correlation peak, the number of secondary peaks, the ratio of the amplitude of the main and secondary peaks, the peak shape broadening characteristics, the peak shape symmetry index, and the main peak energy concentration index. The frequency domain feature extraction module is used to map the autocorrelation function sequence to the frequency domain and extract a set of frequency domain features that can characterize the differences in spectral distribution; the set of frequency domain features includes the number of spectral peaks, center frequency position, spectral spread features, effective bandwidth range, spectral flatness index, and spectral symmetry index. The fuzzy fusion decision module is used to combine the time-domain feature set and the frequency-domain feature set, use fuzzy logic theory to perform fusion calculation on the combined features to obtain a comprehensive decision quantity, and compare the comprehensive decision quantity with a preset decision threshold: when the comprehensive decision quantity is greater than or equal to the decision threshold, it is determined that there is spoofing interference in the current received signal and an alarm signal is output; when the comprehensive decision quantity is less than the decision threshold, it is determined that the current received signal is a normal signal.

Images