Spoofing signal detection method and apparatus based on signal capture, and signal receiver

Abstract

In the signal capture-based fraud signal detection method and device and signal receiver provided in the application, a target correlation peak matrix and a corresponding target picture are determined according to a signal capture result of a to-be-detected received signal; a target Transform model is determined, the target Transform model being obtained by training a picture corresponding to real signal data and a picture corresponding to fraud signal data, the real signal data being intercepted in a first correlation peak matrix, the fraud signal data being obtained in a second correlation peak matrix, the first correlation peak matrix only containing real signal data, and the fraud signal and the real signal being superimposed on each other in the second correlation peak matrix; the target picture is subjected to dimension conversion, and the dimension-converted target picture is input into the target Transform model to obtain a fraud signal detection result of the to-be-detected received signal. In this way, the detection accuracy can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of satellite navigation signal processing, and in particular to a spoofing signal detection method and device based on signal acquisition and a signal receiver. BACKGROUND

[0002] A global navigation satellite system (GNSS) is a system that uses satellite signals to provide position, velocity and timing information in a global range. In GNSS technology, a receiving device receives signals from multiple satellites and uses these signals for navigation. However, in actual application, due to the possibility of spoofing signals, the received signals may contain both real signals and spoofing signals, and two correlation peaks will appear in the correlation operation in the receiver acquisition stage. For spoofing signals with different chip offsets, the correlation peaks are also different.

[0003] Specifically, for spoofing signals with large chip offsets, the correlation peaks of spoofing signals and real signals are completely separated, and the number of correlation peaks can be detected to determine whether spoofing signals are added. However, for spoofing signals with small chip offsets, the correlation peaks of spoofing signals and real signals are superimposed on each other, and still appear as a single peak to the outside, which brings great difficulty to the detection of spoofing signals.

[0004] Currently, for spoofing signal detection technology in the acquisition stage, the solutions mainly include traditional double-peak detection algorithm and method based on traditional deep learning model. The traditional double-peak detection method detects spoofing signals by combining the number of correlation peaks in the acquisition stage and the signal power. However, this method can only detect spoofing signals with a chip offset greater than 2 chips. The spoofing signal detection method based on traditional deep learning model uses a deep learning model to learn the features of the correlation peak matrix, and then realizes spoofing signal detection. However, the deep learning model currently used, such as the CNN model, is a neural network based on traditional convolution structure, and has limited fitting ability for data. The detection accuracy obtained by using this model has room for further improvement.

[0005] In summary, the current spoofing signal detection method for small chip offset still has the problem of low detection accuracy. SUMMARY

[0006] The purpose of the present application is to at least solve one of the above technical defects, in particular, the technical defect that the current spoofing signal detection method for small chip offset still has low detection accuracy.

[0007] In a first aspect, the present application provides a spoofing signal detection method based on signal acquisition, applied to a signal receiver, and the method comprises:

[0008] According to the signal capture result of the to-be-detected received signal, a target correlation peak matrix corresponding to the to-be-detected received signal is determined, and a target picture corresponding to the target correlation peak matrix is determined.

[0009] A target Transformer model is determined, the target Transformer model being obtained by training a picture corresponding to real signal data and a picture corresponding to spoofing signal data, the real signal data being obtained by intercepting a first correlation peak matrix according to a first intercepting range and a plurality of first search steps, the spoofing signal data being obtained by intercepting a second correlation peak matrix according to a second intercepting range and a plurality of second search steps, the first correlation peak matrix only containing real signals, and the second correlation peak matrix containing real signals and spoofing signals superimposed on each other;

[0010] The target picture is subjected to dimension conversion, and the target picture subjected to the dimension conversion is input into the target Transformer model to obtain a spoofing signal detection result of the to-be-detected received signal.

[0011] In one of the embodiments, the step of determining the target picture corresponding to the target correlation peak matrix comprises:

[0012] The target correlation peak matrix is scaled so that the scaled target correlation peak matrix meets a required size for input of the target Transformer model;

[0013] The scaled target correlation peak matrix is converted to obtain the target picture.

[0014] In one of the embodiments, the determination process of the first intercepting range comprises:

[0015] A highest peak signal in the first correlation peak matrix is determined, and a code phase and a carrier frequency of the highest peak signal are determined;

[0016] The code phase of the highest peak signal is taken as a center to offset positive and negative N chips to form a first chip intercepting region;

[0017] The carrier frequency of the highest peak signal is taken as a center to offset positive and negative H hertz to form a first frequency domain intercepting region, and the values of N and M are determined according to an offset between the spoofing signal and the real signal;

[0018] The first intercepting range is determined according to the first chip intercepting region and the first frequency domain intercepting region.

[0019] In one of the embodiments, the determination process of the second intercepting range comprises:

[0020] A carrier frequency and a code phase of the real signal are determined;

[0021] The carrier frequency of the real signal is taken as a center to offset positive and negative K hertz to form a second frequency domain intercepting region.

[0022] centered on the code phase of the real signal, a second chip interception region is formed by shifting plus or minus M chips;

[0023] According to the second frequency domain interception region and the second chip interception region, a second interception range is determined, and the values of K and M are determined according to the offset amount of the spoofing signal and the real signal.

[0024] In one of the embodiments, the training process of the target Transformer model includes:

[0025] The pictures corresponding to the real signal data are divided into a first training set and a first test set, and the pictures corresponding to the spoofing signal data are divided into a second training set and a second test set;

[0026] The first training set and the first test set are assigned a first label value, and the second training set and the second test set are assigned a second label value;

[0027] Based on the first label value and the second label value, the first training set and the second training set are used to train the initial Transformer model, and the first test set and the second test set are used to test the trained initial Transformer model, until the trained initial Transformer model meets the preset training end condition, and the target Transformer model is obtained.

[0028] In one of the embodiments, the step of performing dimension conversion on the target picture includes:

[0029] The target picture is convolved to increase the number of channels of the target picture;

[0030] The increased number of channels is used as the sequence length of the target picture, and the length and width of each channel number are combined as the feature dimension of the target picture.

[0031] In one of the embodiments, the target Transformer model includes two fully connected layers, and the two fully connected layers are connected by using an activation function.

[0032] In a second aspect, the present application provides a spoofing signal detection device based on signal capture, applied to a signal receiver, and the device includes:

[0033] A target picture determination module is configured to determine a target correlation peak matrix corresponding to a to-be-detected received signal according to a signal capture result of the to-be-detected received signal, and determine a target picture corresponding to the target correlation peak matrix.

[0034] The target Transformer model determining module is configured to determine a target Transformer model, the target Transformer model being obtained by training pictures corresponding to real signal data and pictures corresponding to spoofing signal data, the real signal data being obtained by intercepting a first correlation peak matrix according to a first intercept range and using a plurality of first search steps, the spoofing signal data being obtained by intercepting a second correlation peak matrix according to a second intercept range and using a plurality of second search steps, the first correlation peak matrix only containing real signal data, and the second correlation peak matrix containing real signal data and spoofing signal data superimposed on each other.

[0035] The spoofing signal detection module is configured to perform dimension conversion on the target picture, input the target picture after the dimension conversion into the target Transformer model, and obtain a spoofing signal detection result of the to-be-detected received signal.

[0036] In a third aspect, the present application provides a storage medium, the storage medium storing computer readable instructions, the computer readable instructions being executed by one or more processors to cause the one or more processors to perform the steps of the spoofing signal detection method based on signal capture according to any one of the above embodiments.

[0037] In a fourth aspect, the present application provides a signal receiver, including one or more processors and a memory.

[0038] The memory stores computer readable instructions, the computer readable instructions being executed by the one or more processors to perform the steps of the spoofing signal detection method based on signal capture in real time according to any one of the above embodiments.

[0039] From the above technical solutions, it can be seen that the embodiments of the present application have the following advantages:

[0040] In the signal capture-based spoofing signal detection method, device and signal receiver provided in the application, the self-attention mechanism of the Transformer model can capture the dependency relationship between different positions in the input sequence, and can better understand the global structure and relative position information of the signal. Therefore, even in the case of a small offset, the model can find the slight difference between the signals by paying attention to the entire signal sequence. In addition, the multi-head attention mechanism of the Transformer model allows the model to pay attention to different aspects of the input sequence at the same time, thereby better capturing the complex relationships in the sequence, so that the model can make a more sensitive response to the slight changes of the spoofing signal. It can be understood that since the feature distribution of the real signal and the spoofing signal is in different correlation peak matrices, and the spoofing signal is derived from the correlation peak matrix superimposed on each other, and through the respective interception range and search step, more different features can be obtained, thereby helping the Transformer model to better understand the difference and mutual superposition of the real signal and the spoofing signal, and then more accurately detecting. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0042] Figure 1 The correlation peak example diagram of the real signal provided by the embodiments of the present application;

[0043] Figure 2 The correlation peak example diagram of the spoofing signal when the code offset of the spoofing signal pseudo code is 3 chips provided by the embodiments of the present application;

[0044] Figure 3 The correlation peak example diagram of the spoofing signal when the code offset of the spoofing signal pseudo code is 1 chip provided by the embodiments of the present application;

[0045] Figure 4 The flowchart of the signal capture-based spoofing signal detection method provided by the embodiments of the present application;

[0046] Figure 5 The example diagram of the signal capture-based spoofing signal detection method provided by the embodiments of the present application;

[0047] Figure 6 The accuracy effect diagram of the same search step data on different algorithms provided by the embodiments of the present application;

[0048] Figure 7False alarm rate effect diagram of the same search step length data provided by the embodiment of the present application on different algorithms;

[0049] Figure 8 Structure schematic diagram of the spoofing signal detection device based on signal capture provided by the embodiment of the present application;

[0050] Figure 9 Internal structure schematic diagram of the signal receiver provided by the embodiment of the present application;

[0051] Figure 10 Example diagram of the GNSS receiver provided by the embodiment of the present application;

[0052] Figure 11 Detection result example diagram of the TEXBAT data provided by the embodiment of the present application. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0054] The GNSS receiver obtains its own position, speed, time and other information by processing satellite signals. After receiving satellite signals through an antenna, the satellite signals first enter a radio frequency front-end processing, and are down-converted to an intermediate frequency and subjected to analog-to-digital conversion. Then, an estimation value of a carrier frequency and a code phase is obtained through an acquisition stage. Subsequently, a tracking stage is entered, the carrier and pseudo code in the signal are stripped, observation quantities such as carrier phase and pseudo range are obtained, and navigation messages are demodulated. Finally, through a positioning solution module, the receiver obtains a final positioning result.

[0055] After the receiver receives satellite signals through an antenna, the satellite signals first enter an acquisition stage of the signal, and the received navigation signals are correlated with a locally replicated carrier and pseudo-random code. At a position corresponding to the carrier frequency and code phase of the received signal, a correlation peak value is obtained. From this, a preliminary estimation value of the Doppler frequency and code phase is obtained, and a tracking loop is initialized. The correlation peak of the real signal is as shown in Figure 1 .

[0056] When a spoofing signal is added, the signal read by the receiver contains both the real signal and the spoofing signal. Therefore, after the correlation operation in the acquisition stage, two correlation peaks are obtained in the acquisition matrix. As shown in Figure 2As shown, the capture result of the code phase of the spoofing signal offset by 3 chips compared to the real signal is shown, and it can be clearly seen that there are two correlation peaks in the capture matrix. For the spoofing signal with large code chip offset, the correlation peaks of the spoofing signal and the real signal are completely separated, and the number of correlation peaks can be detected to determine whether there is a spoofing signal. However, for the spoofing signal with small code chip offset, the correlation peaks of the spoofing signal and the real signal are superimposed on each other, and still present a single peak to the outside. This brings great difficulty to the detection of the spoofing signal. For example Figure 3 As shown, at this time, the code phase of the spoofing signal is offset by only 1 chip compared to the real signal, and the correlation peaks of the spoofing signal and the real signal are not separated. For the spoofing scene with small code chip offset, the traditional double-peak detection algorithm is invalid. Therefore, for the spoofing signal with small code chip offset, other spoofing detection methods need to be introduced.

[0057] Currently, the solutions for small code chip delay spoofing signals mainly include traditional double-peak detection algorithms and methods based on traditional deep learning models.

[0058] The traditional double-peak detection method combines the number of correlation peaks in the capture stage and the signal power to detect the spoofing signal. In the absence of a spoofing signal, a suitable capture threshold is set, and generally only one correlation peak value greater than the threshold exists; if there is a spoofing signal, there will be multiple correlation peaks greater than the threshold. Therefore, when multiple correlation peaks greater than the threshold are detected, it can be determined that there is a spoofing signal in the current received signal, and at this time, the spoofing signal power is greater than the real signal power, which can be used to detect the spoofing signal. However, the traditional double-peak detection method can only detect spoofing signals with a correlation peak offset of more than 2 chips. When the code chip offset of the spoofing signal is small, the correlation peaks of the spoofing signal and the real signal are superimposed on each other, resulting in a single peak state to the outside. Therefore, the detection method based on the detection of the number of correlation peaks is invalid at this time.

[0059] The spoofing signal detection method based on the traditional deep learning model uses the model to learn the features of the correlation peak matrix, and then realizes the detection of the spoofing signal. Specifically, before and after the spoofing signal is added, the specified range of the receiver capture result is intercepted as a data set, and different label values are given to the correlation peak matrix before and after the spoofing signal is added, and the deep learning model is trained to enhance the model's ability to distinguish whether there is a spoofing signal. Thus, the trained model can be used to detect different spoofing scenarios. Specifically, Li Junzhi et al. introduced the CNN model into the spoofing detection field, and through the training of the CNN model, successfully realized the detection of the spoofing signal with small code chip offset, and obtained good detection effect. However, the CNN model is a neural network based on the traditional convolution structure, and the fitting ability of the data is limited, and the detection accuracy obtained by using the model has room for further improvement.

[0060] Based on this, the application provides a signal capture-based spoofing signal detection method. The following embodiments are described by taking the application of the method in a signal receiver as an example. It can be understood that the signal receiver can be a device with signal receiving and signal processing functions, and can include but not limited to a GNSS receiver, etc. As shown in the figure, the method includes: Figure 4

[0061] S101: According to the signal capture result of the to-be-detected received signal, determine the target correlation peak matrix corresponding to the to-be-detected received signal, and determine the target picture corresponding to the target correlation peak matrix.

[0062] Wherein, the to-be-detected signal refers to the signal that needs to be detected for spoofing signal detection, usually refers to the signal that may exist interference or disguise. The signal capture result refers to the signal processing result obtained by the receiver in the capture stage, including Doppler frequency, code phase and other information. The target correlation peak matrix refers to the correlation peak matrix corresponding to the to-be-detected signal according to the characteristic parameters of the to-be-detected signal. The target picture refers to the image related to the to-be-detected signal in the spoofing signal detection, which is usually converted from the correlation peak matrix.

[0063] In this step, first, analyze the received signal capture result, including Doppler frequency, code phase and other information, which will help to identify the possible to-be-detected signal. Then, according to the information in the signal capture result, extract the correlation peak matrix. The correlation peak matrix is obtained by correlating the received signal with the locally copied carrier and pseudo-random code, and the peak value in it represents the characteristic parameters of the signal, such as Doppler frequency and code phase. Then, according to the characteristic parameters of the to-be-detected signal, compare with the extracted correlation peak matrix to determine the target correlation peak matrix corresponding to it. Once the target correlation peak matrix is determined, it can be converted into a target picture next. Usually, the correlation peak matrix can be represented by the visualization of the matrix, so as to generate the target picture.

[0064] S102: Determine the target Transformer model.

[0065] Wherein, the target Transformer model is obtained by training the picture corresponding to the real signal data and the picture corresponding to the spoofing signal data, the real signal data is obtained by intercepting the first correlation peak matrix according to the first intercepting range and a plurality of first search steps, the spoofing signal data is obtained by intercepting the second correlation peak matrix according to the second intercepting range and a plurality of second search steps, the first correlation peak matrix only contains real signals, and the second correlation peak matrix contains real signals and spoofing signals.

[0066] ​A target Transformer model refers to a deep learning model used for natural language processing (NLP) tasks, which mainly consists of a self-attention mechanism that allows the model to establish dependencies between different positions in a sequence. This mechanism enables the Transformer to process the entire input sequence simultaneously, unlike RNNs which process it step by step. In this application, the target Transformer model is trained to process real signal data and spoof signal data to distinguish between them. Real signal data refers to data samples from real signals, which can be generated by radars, communication systems, or other sensors. Spoof signal data refers to simulated or artificially generated deceptive signal data designed to mimic or disguise as real signals to confuse or deceive the target system. The first and second clipping ranges refer to specific ranges or areas of data selected in the correlation peak matrix. The first clipping range is selected to extract real signal data, while the second clipping range is selected to extract spoof signal data. The first and second search steps refer to the step size or interval when extracting the correlation peak matrix data. A smaller search step can result in more detailed data sampling, while a larger search step can result in more coarse data sampling. The first and second correlation peak matrices refer to the correlation peak matrices extracted from real signals and spoof signals.

[0067] In this step, first, pictures corresponding to real signal data and spoof signal data need to be prepared. These data can come from different sensors or systems, where the pictures corresponding to real signals can be different from those corresponding to spoof signals. For real signals and spoof signals, the first and second correlation peak matrices are extracted, respectively. The correlation peak matrix is commonly used to process signal data and can extract important features in the signal. For the first and second correlation peak matrices, the clipping range and search step need to be set. In the first clipping range, data is clipped in the first correlation peak matrix with multiple first search steps, while in the second clipping range, data is obtained in the second correlation peak matrix with multiple second search steps, to extract different levels of detail information from the signal data. The extracted pictures corresponding to real signal data and pictures corresponding to spoof signal data are used as a training set to train the target Transformer model. The goal of the model is to accurately distinguish between real signals and spoof signals and to learn the characteristics of the signals from the first and second correlation peak matrices.

[0068] The search step is set as a parameter of signal processing inside the receiver. If the search step is set too large, important information in signal acquisition may be lost, resulting in a missed detection. If the search step of the signal is set too small, it will bring huge calculation amount, increase the calculation burden of the receiver, and slow down the processing speed of the receiver.

[0069] It can be understood that, for signal data, the Transformer model can effectively capture the timing information and spatial features in the signal, and can read the complete picture information at one time, has strong data fitting ability, so as to more accurately detect the fraud signal. The first correlation peak matrix only contains the real signal, and the second correlation peak matrix contains the mixture of the two signals, which is to simulate the situation in the actual scene. In actual application, there may be signal superposition, noise interference and the like, therefore, when training the model, providing such mixed signal data helps the model to learn to distinguish different signals and process complex environment. The setting of the interception range and the plurality of search steps is to extract feature information of different scales and levels from the signal data. Different interception ranges and search steps can capture different frequency components and timing patterns in the signal, thereby enriching the input features of the model and improving the representation ability of the model.

[0070] S103: performing dimension conversion on the target picture, inputting the target picture after the dimension conversion into the target Transformer model, and obtaining a fraud signal detection result of the to-be-detected received signal.

[0071] The dimension conversion refers to converting the data format or dimension of the target picture to adapt to the input requirements of the model. For example, it may be necessary to adjust the size of the picture to the input size required by the model, or to perform grayscale, normalization and other preprocessing operations. The fraud signal detection result refers to the judgment result of the model output on whether the to-be-detected received signal is fraudulent, which can be a probability or a classification label, indicating whether the received signal is fraudulent or tampered.

[0072] In this step, the target picture is dimensionally converted to adapt to the input requirements of the target Transformer model. For example, the size of the picture can be adjusted to the input size required by the model, or if the model requires a specific color space or channel number, the picture can be converted from RGB to grayscale or other color spaces, and the pixel values of the image can also be normalized to a value range of 0 to 1 or -1 to 1, so as to better match the model input.

[0073] After that, the target picture after dimension conversion is provided as input to the target Transformer model. The input picture is inferred using the target Transformer model, that is, it is processed and the corresponding output is generated. Finally, the result of the model output is used to explain the fraud signal detection result of the received signal to be detected. Usually, this can involve comparing the probability value or classification label output by the model with a pre-defined threshold to determine whether the received signal is fraudulent or tampered.

[0074] In the above embodiment, the self-attention mechanism possessed by the Transformer model can capture the dependency between different positions in the input sequence, and can better understand the global structure and relative position information of the signal. Therefore, even in the case of small offsets, the model can find the small differences between signals by paying attention to the entire signal sequence. In addition, the multi-head attention mechanism possessed by the Transformer model allows the model to pay attention to different aspects of the input sequence at the same time, thereby better capturing the complex relationships in the sequence, so that the model can be more sensitive to the small changes in the fraudulent signal. It can be understood that since the feature distribution of the real signal and the fraudulent signal is in different correlation peak matrices, and the fraudulent signal is derived from the correlation peak matrix superimposed on each other, and through the respective interception range and search step, more different features can be obtained, thereby helping the Transformer model to better understand the differences and mutual superposition of the real signal and the fraudulent signal, and thus more accurately detect.

[0075] In one embodiment, the step of determining the target picture corresponding to the target correlation peak matrix comprises:

[0076] scaling the target correlation peak matrix so that the scaled target correlation peak matrix meets the required size for input of the target Transformer model;

[0077] converting the scaled target correlation peak matrix to obtain the target picture.

[0078] Specifically, first, the size required for the input of the target Transformer model is determined, which can be a fixed size or certain input requirements of the model, such as specific dimensions of width and height, or the size corresponding to the dataset used when training the Transformer model. Then, the target correlation peak matrix is scaled to the required size using appropriate scaling algorithms, such as bilinear interpolation, nearest neighbor interpolation, etc. This can be done through image processing libraries such as OpenCV, PIL, etc. The scaled correlation peak matrix is converted to image format, which can be achieved by mapping the values in the matrix to the range of pixel values, for example, normalizing the values in the matrix to between 0 and 255 and treating them as pixel values of the image. Further, if the values in the correlation peak matrix represent the intensity or grayscale values of the image, they can be directly used as the pixel values of the generated image. If the values in the correlation peak matrix need to be further processed, such as normalization, color mapping, etc., appropriate processing can be performed according to the specific situation.

[0079] It can be understood that in actual processing, there may be cases where the target correlation peak is too large or too small, so scaling and converting the target correlation peak matrix to meet the input requirements of the Transformer model can better utilize the advantages of the model, improve processing efficiency and accuracy, and thus better achieve analysis and processing of the correlation peak matrix.

[0080] In one embodiment, the determination process of the first interception range includes:

[0081] Determine the highest peak signal in the first correlation peak matrix, and determine the code phase and carrier frequency of the highest peak signal;

[0082] Centering on the code phase of the highest peak signal, offset by plus and minus N chips to form a first chip interception area;

[0083] Centering on the carrier frequency of the highest peak signal, offset by plus and minus H Hertz to form a first frequency domain interception area, the values of N and M are determined according to the offset between the spoofing signal and the real signal;

[0084] Determine the first interception range according to the first chip interception area and the first frequency domain interception area.

[0085] Wherein, the highest peak in the correlation peak matrix represents the main component of the signal, which is usually used to extract the key information of the signal. In digital communication, signals are usually divided into different chips, and code phase represents the phase offset of each chip. The carrier frequency of the signal refers to the basic frequency component in the signal, which determines the center frequency of the signal. Chip is a basic signal unit in digital communication, used to represent digital information. Frequency domain represents the characteristics of the signal in frequency. The interception area refers to the area or range of the selected part of interest.

[0086] Specifically, the highest peak in the correlation peak matrix is found by a correlation analysis method, and parameters such as the phase and carrier frequency thereof are obtained. Taking the code phase of the highest peak signal as the center, the first code chip interception region is formed by shifting the positive and negative N code chips. Taking the carrier frequency of the highest peak signal as the center, the first frequency domain interception range is formed by shifting the positive and negative H hertz. The first interception range is determined by comprehensively considering the code chip axis interception region and the frequency domain interception range.

[0087] Further, for the selected first code chip interception region, a plurality of corresponding search step lengths can be determined, and for the selected first frequency domain interception region, a plurality of corresponding search step lengths can be determined, and the two search step lengths are combined into different search step length combinations to obtain real signal data in the first interception range. In one example, taking the highest peak in the correlation peak matrix as the center, a region of ±2 code chips is intercepted on the code chip axis, and a range of ±2 kHz is intercepted on the carrier frequency axis. The search step lengths of the code chips are set to half a code chip and one quarter of a code chip, respectively, and the search step lengths of the frequency domain are set to 500 Hz and 250 Hz, respectively. Thus, a total of four different search step length combination data can be obtained, which are: (1) code chip axis: 0.5 code chip, frequency domain axis: 500 Hz, (2) code chip axis: 0.25 code chip, frequency domain axis: 500 Hz, (3) code chip axis: 0.5 code chip, frequency domain axis: 250 Hz, and (4) code chip axis: 0.25 code chip, frequency domain axis: 250 Hz.

[0088] It can be understood that the values of N and M are determined according to the offset between the spoof signal and the real signal. For a spoof signal with a small code chip offset, the code phase corresponding to the spoof signal generally does not differ from the real signal by more than 2 code chips, and the Doppler shift of the spoof signal generally does not differ from the real signal by more than 2 kHz. Therefore, when the spoof signal exists, the correlation peaks of the spoof signal and the real signal can generally be seen simultaneously within this range. When the spoof signal does not exist, only the real signal exists within this range. Based on the difference between the two, a Transformer model is used for classification. In addition, if the code phase of the spoof signal is offset from the real signal by more than 2 code chips, the interception range on the code chip should be expanded so that the interception result contains both the spoof signal and the real signal. The processing method on the carrier frequency axis is similar. In addition, if the offset of the spoof signal in the code phase and the carrier frequency from the real signal is less than 2 code chips and 2 kHz, the corresponding interception range can be correspondingly reduced.

[0089] In this embodiment, by determining the highest peak signal in the first correlation peak matrix, the main characteristics of the signal such as code phase and carrier frequency can be accurately determined, which helps to extract key information from complex signals. Taking the code phase of the highest peak signal as the center, shifting by plus or minus N chips, the first chip interception region can be determined to accurately locate the position of the signal in the time domain, making the subsequent processing more accurate. Taking the carrier frequency of the highest peak signal as the center, shifting by plus or minus H Hz, the first frequency domain interception region is formed, which helps to accurately locate the signal in the frequency domain and determine its frequency range. By considering the chip axis interception region and the frequency domain interception region, the first interception range is determined, which can provide a comprehensive signal interception range to ensure the accuracy and reliability of subsequent processing and analysis. By accurately intercepting the signal, the amount and complexity of processing can be reduced, and the efficiency and performance of subsequent processing can be improved.

[0090] In one embodiment, the determination process of the second interception range includes:

[0091] determining the carrier frequency and code phase of the real signal;

[0092] forming a second frequency domain interception region by shifting plus or minus K Hz from the carrier frequency of the real signal;

[0093] forming a second chip interception region by shifting plus or minus M chips from the code phase of the real signal;

[0094] determining the second interception range according to the second frequency domain interception region and the second chip interception region, and the values of K and M are determined according to the offset of the spoof signal and the real signal.

[0095] Specifically, signal processing techniques such as autocorrelation function, correlation peak detection, etc. can be used to determine the carrier frequency and code phase of the real signal. Taking the carrier frequency of the real signal as the center, shifting by plus or minus K Hz, the second frequency domain interception region can be formed, which can be achieved by spectral analysis or frequency domain filtering, etc. Taking the code phase of the real signal as the center, shifting by plus or minus M chips, the second chip interception region can be formed, which can be achieved by chip synchronization technology or chip edge detection, etc. According to the second frequency domain interception region and the second chip interception region, the second interception range can be determined by merging or crossing the frequency domain interception region and the chip interception region to ensure that the complete information of the real signal is included.

[0096] Further, in actual spoofing scenarios, whether it is a repeater spoofing or a generative spoofing, the power of the spoofing signal is greater than that of the real signal, so when the receiver performs signal acquisition, according to the height of the correlation peak, that is, the power of the signal, the receiver will capture the correlation peak corresponding to the spoofing signal, thereby tracking the spoofing signal, and finally obtaining the positioning result corresponding to the spoofing signal. Based on this, the power of the spoofing signal can be set to 1.1-1.3 times the power of the real signal, and the specific size is randomly selected. Both can ensure the power advantage of the spoofing signal relative to the real signal, and can increase the diversity of the spoofing signal, which is more representative for actual spoofing scenarios, making the detection result more convincing. In one example, the power of the spoofing signal is set to 1.1-1.3 times the amplitude of the power of the real signal. In addition, the Doppler shift of the spoofing signal is centered on the carrier frequency of the real signal, and is randomly selected within the range of ±1 kHz near the carrier frequency. Centered on the code phase of the real signal, the code phase of the spoofing signal starts from a shift of 0.1 chip and goes up to 2 chips to obtain 20 types of data with different chip shift sizes.

[0097] In this embodiment, the second clipping range is determined based on the code phase and the carrier frequency of the real signal, so that the spoofing signal data with small shift can be more accurately obtained.

[0098] In one embodiment, the training process of the target Transformer model includes:

[0099] The pictures corresponding to the real signal data are divided into a first training set and a first test set, and the pictures corresponding to the spoofing signal data are divided into a second training set and a second test set;

[0100] The first training set and the first test set are assigned a first label value, and the second training set and the second test set are assigned a second label value;

[0101] Based on the first label value and the second label value, the first training set and the second training set are used to train the initial Transformer model, and the first test set and the second test set are used to test the trained initial Transformer model, until the trained initial Transformer model meets the preset training end condition, and the target Transformer model is obtained.

[0102] Specifically, the real signal pictures and the spoof signal pictures are divided into a first training set, a first test set, a second training set, and a second test set, respectively. The first training set and the first test set are assigned a first label value, for example, real signal is label 1. The second training set and the second test set are assigned a second label value, for example, spoof signal is label 2. Based on the first label value and the second label value, the initial Transformer model is trained using the first training set and the second training set. During the training process, a common Transformer model structure can be used, and adjustments and optimizations can be made according to actual conditions. An appropriate loss function (such as cross-entropy loss) can also be used to measure the difference between the model output and the true label, and to optimize the model parameters. The trained initial Transformer model is tested using the first test set and the second test set. The performance of the model on the test set can be evaluated, including accuracy, precision, recall, and other indicators. The model is adjusted and optimized according to the test results to improve its performance. A preset training end condition is set, such as reaching a specified number of training rounds, model convergence, etc. When the model meets the preset end condition, the training process ends, and the target Transformer model is obtained.

[0103] In this embodiment, training is performed using real signals and spoof signals respectively, which can enable the model to learn more rich feature representations, thereby improving its generalization ability and making its performance on new data more robust. At the same time, spoof signals usually have large variations and noise, and by using these data in training, the model can have better adaptability to various noise and variations, improving the robustness of the model. In addition, after the target Transformer model learns the features of both real signals and spoof signals, it will be more effective in distinguishing between the two, thereby improving the detection and recognition ability of spoof signals.

[0104] In one embodiment, the step of performing dimension conversion on the target picture includes:

[0105] The target picture is convolved to increase the number of channels of the target picture.

[0106] The increased number of channels is used as the sequence length of the target picture, and the length and width of each channel number are combined as the feature dimension of the target picture.

[0107] Specifically, the target picture is subjected to a convolution operation, and in the process of the convolution operation, the number of convolution kernels can be set to increase the channel number of the target picture. Each convolution kernel corresponds to a channel, and by adjusting the number of convolution kernels, the channel number of the target picture can be increased, thereby increasing the information representation capability of the picture. For example, a convolution with a convolution kernel size of 1*1 can be performed. The increased channel number is taken as the sequence length of the target picture, meaning that the picture is regarded as a sequence, with each channel corresponding to a time step in the sequence. The purpose of this is to organize the information of the picture into a sequence form for subsequent input into a Transformer model for processing. For each channel number, the height and width can be combined into a feature dimension. For example, if the size of the target picture after the convolution operation is (H, W, C), where H represents the height, W represents the width, and C represents the channel number, it can be transformed into a tensor with a shape of (C, HW), where each channel corresponds to a feature vector with a length of HW. In this way, the target picture can be converted into the required dimension of the Transformer model for normal processing.

[0108] In this embodiment, the convolution operation can extract various features in the picture, and increasing the channel number can increase the diversity and richness of the features, enabling the model to capture more rich picture feature information. Combining the height and width of each channel number into a feature dimension is equivalent to integrating the feature map information of each channel into a vector, reducing the loss of information and helping to retain more important features. After increasing the channel number and the feature dimension, the model can better understand and represent the features of the input picture, thereby improving the performance of the model.

[0109] In one embodiment, the target Transformer model includes two fully connected layers connected by an activation function.

[0110] The fully connected layer is a layer type in neural networks, also known as a dense connection layer or an affine layer. Each neuron of the fully connected layer is connected to all the neurons of the previous layer, and each connection has a corresponding weight. The role of the fully connected layer is to perform linear transformation on the input features and capture complex relationships between features by learning the weights. The activation function is a non-linear transformation between fully connected layers, which introduces non-linear factors into the network, enabling the neural network to learn and represent more complex function relationships. Common activation functions include ReLU (Rectified Linear Unit), Sigmoid, Tanh, etc., each with different characteristics and application scenarios.

[0111] Further, after the data is processed by the Transformer model, the data includes three dimensions: batch size, sequence length, and feature dimension. However, the full connection layer requires that the data has only two dimensions, i.e., batch size and data feature. Therefore, in order to process the data in the full connection layer, the last two dimensions of the data need to be combined, and the data needs to be converted into two dimensions, so that the full connection layer can normally process the data.

[0112] To facilitate understanding of the scheme of the present application, a specific example is provided below for illustration.

[0113] As shown in Figure 5 , using the correlation peak matrix obtained by the navigation receiver in the acquisition stage, it is first determined whether the correlation peak value exceeds the acquisition threshold. For the correlation peak greater than the threshold, a specified range is intercepted, and the intercepted result is converted into a picture output. The code phase offset size of the spoofing signal is adjusted to obtain spoofing signals with different chip delays. The spoofing signals are combined with the real signals, and different labels are given respectively to obtain a complete data set. For the obtained data set, the training data and the test data are randomly divided, the training data is used to train the parameters of the deep learning model built, and the learning ability of the model for the different features of the acquisition matrix before and after the addition of the spoofing signal is improved. The test data is input into the trained model to obtain the output result, and the actual detection ability of the model for the spoofing signal is tested.

[0114] In order to verify the detection effect of the model proposed in the present method, MLP, CNN and LSTM models are used as comparison algorithms. For the simulated data set, the detection effects of the four types of models are shown in Figure 5 . As can be seen from the figure, with the increase of the chip offset, the detection accuracy of the data set with different search steps is continuously increasing. When the chip delay of the spoofing signal is greater than 0.5 chips, the detection accuracy tends to 100%. From the detection values, the detection accuracy based on the Transformer model is the highest, and for the data of various search steps, the detection accuracy is more than 95%. For the data with a search step of 0.5 chips, when the minimum chip delay is 0.1 chip, the detection accuracy can reach 97%, which is obviously better than the detection effects of the other models. In summary, for data with different search steps, the detection effect of the Transformer model is better than that of the other models. The detection result based on the LSTM model is better than that of the CNN model. The detection effect of the MLP is the worst.

[0115] The false alarm rates of various models on the same group of data are shown in Figure 7As shown, it can be seen that the false alarm rate of the Transformer model is the lowest, and the highest is not more than 4%, and the false alarm rate of part of the data is at most 2%. The false alarm rate of other algorithms decreases similarly to the Transformer model, but the specific false alarm rate value is higher than that of the Transformer model. Especially for the MLP model, for the data with an offset of 0.1 chip, the false alarm rate value is as high as nearly 6%, and the descending speed of the false alarm rate curve is also slower than that of the Transformer model. Therefore, for data with different search steps, the false alarm rate based on the Transformer model is the lowest. The false alarm rate of the MLP model is the highest. The false alarm rates of the CNN and LSTM algorithms are in the middle.

[0116] The signal capture-based spoofing signal detection device provided in the embodiments of the present application is described below. The signal capture-based spoofing signal detection device described below can be correspondingly referred to the signal capture-based spoofing signal detection method described above. As shown in Figure 8 The present application provides a signal capture-based spoofing signal detection device, applied to a signal receiver, and the device comprises:

[0117] A target picture determination module 201 is configured to determine a target correlation peak matrix corresponding to a to-be-detected received signal according to a signal capture result of the to-be-detected received signal, and determine a target picture corresponding to the target correlation peak matrix.

[0118] A target Transformer model determination module 202 is configured to determine a target Transformer model, which is obtained by training a picture corresponding to real signal data and a picture corresponding to spoofing signal data, the real signal data is obtained by intercepting a first correlation peak matrix according to a first intercepting range and a plurality of first search steps, the spoofing signal data is obtained by intercepting a second correlation peak matrix according to a second intercepting range and a plurality of second search steps, the first correlation peak matrix only contains real signals, and the second correlation peak matrix contains real signals and spoofing signals.

[0119] A spoofing signal detection module 203 is configured to perform dimension conversion on the target picture, input the target picture after the dimension conversion into the target Transformer model, and obtain a spoofing signal detection result of the to-be-detected received signal.

[0120] In one embodiment, the target picture determination module 201 comprises:

[0121] A target correlation peak matrix scaling unit is configured to scale the target correlation peak matrix, so that the scaled target correlation peak matrix meets the required size for inputting the target Transformer model;

[0122] The target picture determination unit is configured to convert the scaled target correlation peak matrix to obtain a target picture.

[0123] In one embodiment, the target Transformer model determination module 202 includes:

[0124] The highest peak signal determination unit is configured to determine a highest peak signal in the first correlation peak matrix, and determine a code phase and a carrier frequency of the highest peak signal.

[0125] The first chip interception region forming unit is configured to form a first chip interception region with the code phase of the highest peak signal as the center and offset by plus and minus N chips.

[0126] The first frequency domain interception region forming unit is configured to form a first frequency domain interception region with the carrier frequency of the highest peak signal as the center and offset by plus and minus H hertz, and the values of N and M are determined according to the offset between the spoofing signal and the real signal.

[0127] The first interception range determination unit is configured to determine a first interception range according to the first chip interception region and the first frequency domain interception region.

[0128] In one embodiment, the target Transformer model determination module 202 includes:

[0129] The code phase determination unit is configured to determine a carrier frequency and a code phase of the real signal.

[0130] The second frequency domain interception region forming unit is configured to form a second frequency domain interception region with the carrier frequency of the real signal as the center and offset by plus and minus K hertz.

[0131] The second chip interception region forming unit is configured to form a second chip interception region with the code phase of the real signal as the center and offset by plus and minus M chips.

[0132] The second interception range determination unit is configured to determine a second interception range according to the second frequency domain interception region and the second chip interception region, and the values of K and M are determined according to the offset between the spoofing signal and the real signal.

[0133] In one embodiment, the target Transformer model determination module 202 includes:

[0134] The training set division unit is configured to divide the pictures corresponding to the real signal data into a first training set and a first test set, and divide the pictures corresponding to the spoofing signal data into a second training set and a second test set.

[0135] The label value assignment unit is configured to assign a first label value to the first training set and the first test set, and assign a second label value to the second training set and the second test set.

[0136] The model training unit is configured to train the initial Transformer model based on the first label value and the second label value, using the first training set and the second training set, and test the trained initial Transformer model based on the first test set and the second test set, until the trained initial Transformer model meets a preset training end condition, and obtain a target Transformer model.

[0137] In one embodiment, the spoofing signal detection module 203 comprises:

[0138] The target picture convolution unit is configured to convolve the target picture to increase the number of channels of the target picture.

[0139] The feature dimension determination unit is configured to take the increased number of channels as the sequence length of the target picture, and take the length and width of each channel number combined as the feature dimension of the target picture.

[0140] In one embodiment, the target Transformer model comprises two fully connected layers connected by an activation function.

[0141] In one embodiment, the present application further provides a storage medium having computer readable instructions stored therein, and the computer readable instructions, when executed by one or more processors, cause the one or more processors to perform the steps of the spoofing signal detection method based on signal capture as described in any of the above embodiments.

[0142] Traditional detection methods are all tested on simulation data, and the detection method in the capture stage is not applied to the actual navigation spoofing signal detection task. Based on this, in one embodiment, the present application further provides a signal receiver having computer readable instructions stored therein, and the computer readable instructions, when executed by one or more processors, cause the one or more processors to perform the steps of the spoofing signal detection method based on signal capture in real time as described in any of the above embodiments. In this way, the spoofing signal detection method based on signal capture can be applied to practical applications, thereby improving the practicability of the method.

[0143] As shown in Figure 9 , Figure 9 FIG. 3 shows an internal structure diagram of a signal receiver according to an embodiment of the present application. The signal receiver 300 can be provided as a server. Referring to Figure 9The signal receiver 300 comprises a processing component 302, which further comprises one or more processors, and a memory resource represented by the memory 301 for storing instructions executable by the processing component 302, such as an application program. The application program stored in the memory 301 can comprise one or more than one module each corresponding to a set of instructions. In addition, the processing component 302 is configured to execute the instructions to perform the signal capture based spoofing signal detection method of any of the embodiments described above.

[0144] The signal receiver 300 can further comprise a power supply component 303 configured to perform power management of the signal receiver 300, a wired or wireless network interface 304 configured to connect the signal receiver 300 to a network, and an input / output (I / O) interface 305. The signal receiver 300 can operate based on an operating system stored in the memory 301, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM or the like.

[0145] Those skilled in the art can understand that, Figure 9 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the signal receiver to which the scheme of the present application is applied. The specific signal receiver can comprise more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0146] In one example, as Figure 10 In order to further verify the detection effect of the Transformer model, the detection model is embedded into an actual receiver, and the model detection capability is verified by using actually collected data.

[0147] FGI-GSRx is a GNSS software receiver developed based on Matlab. The application can embed the detection model of the Transformer in the FGI-GSRx receiver, so that the receiver can detect spoofing signals in the process of signal processing. The modified receiver only leaves the code of the signal capture stage. For the read navigation signal, determine the reading time of the signal, enter the capture stage of the signal. When the correlation peak of the capture is greater than the threshold, the capture matrix of this channel is intercepted, the intercepted matrix is converted into a picture output, and then sent to the Transformer model embedded in the receiver to obtain the discrimination result of the model whether it is a spoofing signal. Then judge the time when the receiver reads the signal, if it has not reached the time endpoint, add a time step to the original time reference, read the signal again at the next time point, and perform capture operation and spoofing signal detection on the capture result. In this way, continuous signal capture and spoofing signal detection can be performed on the entire data.

[0148] The TEXBAT dataset is an open-source intermediate spoofing interference dataset, which contains 8 spoofing scenarios. The spoofing interference used is based on the GPS L1 signal, with a sampling rate of 25MHz. The application can use the capture matrix of the signal channel with prn number 16 in scenario 3 and scenario 4 as a dataset to train the Transformer model. Then use the trained model to detect the signal channel with prn number 13 in the corresponding scenario. The time range for reading the signal is set to 0-410s, and the time step increased each iteration is 1s. The Transformer model is used for spoofing signal detection. The detection result is as shown in Figure 11

[0149] As can be seen from Figure 11 , the output of the detection result has three cases in total. The output of 0 indicates that the Transformer model considers that only real signals exist at this time, the output of 1 indicates that the Transformer model judges that spoofing signals are added at this time, and the output of 0.5 indicates that no satellite of the specified channel is captured in this round of signal capture.

[0150] ​In the first 100s of the signal, there is only real signal. When the spoofing signal starts to join around 100s, the code phase of the spoofing signal is aligned with the real signal, and the top view of the capture matrix does not change. So the model will always judge it as a real signal before the correlation peak of the spoofing signal starts to deviate. Until close to 200s, the correlation peak of the spoofing signal starts to separate from the real signal in code phase, and the capture matrix starts to change significantly. At this time, the Transformer model can identify the changes in the capture matrix and give a spoofing alarm.

[0151] Finally, it should be noted that in this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element. "A," "an," and "the" are also used herein to include one or more instances of (e.g., the phrase "the element" as used herein should be interpreted to cover one or more instances of that element). The term "plurality" is intended to mean one or more than one, i.e., at least two. The term "and / or" includes any and all combinations of one or more of the associated listed items.

[0152] The various embodiments described in this specification are presented as examples of the application. Each embodiment is presented highlighting the differences from the other embodiments, and the various embodiments can be combined as desired, and the same and similar parts are cross-referenced.

[0153] The above description of disclosed embodiments provides enabling disclosure sufficient for one of ordinary skill in the art to implement or use the application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A signal capture based spoofing signal detection method, characterized in that, The method is applied to a signal receiver, and the method comprises the following steps: According to the signal acquisition result of the to-be-detected received signal, a target correlation peak matrix corresponding to the to-be-detected received signal is determined, and a target picture corresponding to the target correlation peak matrix is determined; A target Transformer model is determined, the target Transformer model being obtained by training a picture corresponding to real signal data and a picture corresponding to spoofing signal data, the real signal data being obtained by intercepting a first correlation peak matrix according to a first intercepting range and using a plurality of first search steps, the spoofing signal data being obtained by intercepting a second correlation peak matrix according to a second intercepting range and using a plurality of second search steps, the first correlation peak matrix only containing real signals, and the second correlation peak matrix containing real signals and spoofing signals superimposed on each other; The target picture is subjected to dimension conversion, and the target picture subjected to dimension conversion is input into the target Transformer model to obtain a spoofing signal detection result of the to-be-detected received signal.

2. The signal capture based spoofing signal detection method of claim 1, wherein, The step of determining the target picture corresponding to the target correlation peak matrix comprises the following steps: The target correlation peak matrix is scaled so that the scaled target correlation peak matrix meets the required size for inputting into the target Transformer model; The scaled target correlation peak matrix is converted to obtain the target picture.

3. The signal capture based spoofing signal detection method of claim 1, wherein, The determination process of the first intercepting range comprises the following steps: The highest peak signal in the first correlation peak matrix is determined, and the code phase and carrier frequency of the highest peak signal are determined; A first code chip intercepting region is formed by offsetting positive and negative N code chips with the code phase of the highest peak signal as the center; A first frequency domain intercepting region is formed by offsetting positive and negative H hertz with the carrier frequency of the highest peak signal as the center, and the values of N and M are determined according to the offset amount between the spoofing signal and the real signal; The first intercepting range is determined according to the first code chip intercepting region and the first frequency domain intercepting region.

4. The signal capture based spoofing signal detection method of claim 1, wherein, The determination process of the second intercepting range comprises the following steps: The carrier frequency and code phase of the real signal are determined; A second frequency domain intercepting region is formed by offsetting positive and negative K hertz with the carrier frequency of the real signal as the center; A second code chip intercepting region is formed by offsetting positive and negative M code chips with the code phase of the real signal as the center; The second intercepting range is determined according to the second frequency domain intercepting region and the second code chip intercepting region, and the values of K and M are determined according to the offset amount between the spoofing signal and the real signal.

5. The signal capture based spoofing signal detection method of claim 1, wherein, The training process of the target Transformer model comprises the following steps: The picture corresponding to the real signal data is divided into a first training set and a first test set, and the picture corresponding to the spoofing signal data is divided into a second training set and a second test set; The first training set and the first test set are assigned a first label value, and the second training set and the second test set are assigned a second label value; Based on the first label value and the second label value, the first training set and the second training set are used to train an initial Transformer model, and the first test set and the second test set are used to test the trained initial Transformer model until the trained initial Transformer model meets a preset training end condition, and the target Transformer model is obtained.

6. The signal capture based spoofing signal detection method of claim 1, wherein, The step of performing dimension conversion on the target picture comprises: The target picture is convolved to increase the number of channels of the target picture; The increased number of channels is taken as the sequence length of the target picture, and the length and width of each channel number are combined as the feature dimension of the target picture.

7. The signal capture based spoofing signal detection method according to any one of claims 1 to 6, characterized in that, The target Transformer model comprises two fully connected layers, and the two fully connected layers are connected by an activation function.

8. A signal capture based spoofing signal detection apparatus, characterized by, The device is applied to a signal receiver and comprises: A target picture determination module is configured to determine a target correlation peak matrix corresponding to a to-be-detected received signal according to a signal capture result of the to-be-detected received signal, and determine a target picture corresponding to the target correlation peak matrix. A target Transformer model determination module is configured to determine a target Transformer model, wherein the target Transformer model is obtained by training a picture corresponding to real signal data and a picture corresponding to spoofing signal data, the real signal data is obtained by intercepting a first correlation peak matrix according to a first intercepting range and a plurality of first search steps, the spoofing signal data is obtained by intercepting a second correlation peak matrix according to a second intercepting range and a plurality of second search steps, the first correlation peak matrix only contains real signals, and the second correlation peak matrix contains real signals and spoofing signals. A spoofing signal detection module is configured to perform dimension conversion on the target picture, input the dimension-converted target picture into the target Transformer model, and obtain a spoofing signal detection result of the to-be-detected received signal.

9. A storage medium characterized by: The storage medium stores computer readable instructions, and the computer readable instructions are executed by one or more processors to cause the one or more processors to perform the steps of the spoofing signal detection method based on signal capture according to any one of claims 1 to 7.

10. A signal receiver, characterized by Comprise: One or more processors and a memory; The memory stores computer readable instructions, and the computer readable instructions are executed by the one or more processors to perform the steps of the spoofing signal detection method based on signal capture in real time according to any one of claims 1 to 7.

Images