Single-antenna GNSS spoofing interference detection system and method based on CGAN

Abstract

The present invention provides a single-antenna GNSS deception interference detection system and method based on CGAN, which relates to the field of GNSS technology. GNSS navigation signals are received through antennas, and the signals are down-converted into digital intermediate frequency signals; the digital intermediate frequency signals are captured and tracked, and the leading, immediate and lagging correlation function values ​​of the IQ dual-path correlator are extracted; characteristic parameters are extracted from the correlation function values ​​and standardized to generate a training data set; the training data set is enhanced using CGAN to generate a pseudo data set, and the pseudo data set is merged with the original data set into an enhanced data set; the enhanced data set is input into an artificial neural network for training, and the classification result of the deception state is output. The advantages of the generative adversarial network are used to improve the detection ability of the network model for the deception state of the target receiver, and the selected characteristic parameters are mainly for the parameters in the baseband signal processing stage, achieving a more efficient and practical anti-interference technical effect.

Description

Technical Field

[0001] The present application relates to the field of GNSS technology, and in particular to a single-antenna GNSS spoofing interference detection system and method based on CGAN. Background Art

[0002] At the current stage of development, satellite-based navigation systems are one of the most important means of positioning and timing. GNSS can provide all-weather position, velocity and time (PVT) services around the world and is widely used in scientific research, geodesy, aerospace and other fields. However, due to the openness of GNSS signals and their low signal receiving power, many security issues have been exposed. GNSS signals are susceptible to a series of intentional interference, and with the popularization of programmable simulators and software radios, civilian GNSS receivers are also extremely susceptible to interference from low-cost deception devices.

[0003] In order to ensure the service quality of GNSS in complex electromagnetic environments, anti-spoofing and interference technology has become particularly important. At present, there have been a lot of studies on the anti-interference of GNSS receivers. Among them, the deception detection technology based on baseband signal processing has attracted widespread attention from researchers as the most efficient and economical anti-interference solution. Traditional deception detection methods include signal quality monitoring technology (SQM) and power monitoring methods, but the characteristic distribution changes of the known signal are required, and according to statistical principles, the determination of signal models, prior probabilities or likelihood functions is also crucial. In contrast, methods based on machine learning or deep learning do not require specific determination of mathematical models, noise models or their statistical characteristics. Therefore, deception detection methods based on artificial intelligence technology have received widespread attention in the academic community and have made breakthrough progress in feasibility verification.

[0004] However, in order to improve the detection effect of artificial intelligence deception detection technology, the existing technology mainly optimizes the selection of parameter features, but there are still many and complex shortcomings in the selection of specific feature sets, and there is no improvement on the network model. The generalization ability of the network still needs to be improved.

[0005] Therefore, how to provide a more efficient and practical anti-interference solution has become a technical problem that needs to be solved urgently. Summary of the invention

[0006] In order to provide a more efficient and practical anti-interference solution, the present invention provides a single-antenna GNSS spoofing interference detection system and method based on CGAN.

[0007] In the first aspect, the present application provides a single-antenna GNSS spoofing interference detection system based on CGAN, which adopts the following technical solutions:

[0008] A single-antenna GNSS spoofing jamming detection system based on CGAN, comprising:

[0009] A front-end processing module, used for receiving GNSS navigation signals through an antenna and down-converting the navigation signals into digital intermediate frequency signals;

[0010] A baseband signal processing module, connected to the front-end processing module, is used to capture and track the digital intermediate frequency signal, generate a local carrier and a pseudo code signal, and extract the leading, immediate and lagging correlation function values ​​of the IQ dual-path correlator;

[0011] A data set processing module, connected to the baseband signal processing module, for extracting characteristic parameters from the correlation function values, and performing standardization processing on the characteristic parameters to generate a training data set;

[0012] The CGAN-ANN network module is connected to the data set processing module and includes a conditional generative adversarial network and an artificial neural network, wherein the conditional generative adversarial network is used to enhance the training data set to generate a pseudo data set, and the artificial neural network is used to perform deception state detection based on the enhanced data set and output a judgment result of deception or non-deception.

[0013] Optionally, the characteristic parameter includes at least one of the following:

[0014] The received power is determined by calculating the average power of the RF output signal within a preset time window;

[0015] Tracking loop related parameters, including phase code offset and energy loss ratio;

[0016] Statistical characteristics based on moving average and moving variance.

[0017] Optionally, the CGAN includes:

[0018] The generator receives the noise vector and the conditional label as input and generates pseudo samples that are consistent with the real data distribution;

[0019] A discriminator, used to distinguish the pseudo samples from real samples, and optimize network parameters through adversarial training with the generator;

[0020] Among them, the hidden layer of the generator adopts the ReLU activation function, and the output layer adopts the Tanh activation function; the hidden layer of the discriminator adopts the LeakyReLU activation function, and the output layer adopts the Sigmoid activation function.

[0021] Optionally, the artificial neural network includes an input layer, at least two fully connected hidden layers and an output layer, wherein:

[0022] The input layer receives the standardized feature parameters;

[0023] The hidden layer adopts ReLU activation function;

[0024] The output layer adopts a Sigmoid activation function, and the output value is a deception probability between 0 and 1. When the output value is greater than 0.5, it is determined to be a deception state, otherwise it is a non-deception state.

[0025] Optionally, the data set processing module is further used to label the multi-scenario deception signals in the TEXBAT data set, where the normal state is labeled as 0 and the deception state is labeled as 1.

[0026] Optionally, the system further includes a verification module, which is used to verify the deception detection performance of unknown scenarios by using a leave-one-out method, specifically including:

[0027] The data of any scenario is selected as the validation set, and the remaining scenario data is divided into training set and test set in proportion;

[0028] The CGAN-ANN network module is trained through a training set and a test set, and the generalization ability of the model is evaluated through a validation set.

[0029] Optionally, the baseband signal processing module outputs six correlation function values ​​through an IQ dual-path correlator, including leading, immediate and lagging correlation values ​​of the I path and the Q path.

[0030] In a second aspect, the present application provides a single-antenna GNSS spoofing interference detection method based on CGAN, comprising:

[0031] S1: Receive GNSS navigation signals through an antenna and down-convert the signals into digital intermediate frequency signals;

[0032] S2: Capture and track the digital intermediate frequency signal, and extract the leading, immediate and delayed correlation function values ​​of the IQ dual-path correlator;

[0033] S3: extracting characteristic parameters from the correlation function values ​​and performing standardization processing to generate a training data set;

[0034] S4: using CGAN to enhance the training dataset to generate a pseudo dataset, and merging the pseudo dataset with the original dataset into an enhanced dataset;

[0035] S5: Input the enhanced data set into an artificial neural network for training, and output a classification result of the deception state.

[0036] Optionally, in step S4, the generator of the CGAN receives the noise vector and the conditional label to generate a pseudo sample, and the discriminator optimizes the parameters of the generator and the discriminator through adversarial training until the distribution of the generated samples is consistent with that of the real samples.

[0037] Optionally, in step S5, a Sigmoid function is used as an output layer activation function of the ANN, and when the output value is greater than 0.5, it is determined to be a deceptive state, otherwise it is a non-deceptive state.

[0038] In summary, the present application includes the following beneficial technical effects:

[0039] This application receives GNSS navigation signals through antennas and down-converts the signals into digital intermediate frequency signals; captures and tracks digital intermediate frequency signals, extracts the leading, immediate, and lagging correlation function values ​​of the IQ dual-path correlator; extracts characteristic parameters from the correlation function values ​​and performs standardization processing to generate a training data set; uses CGAN to enhance the training data set, generates a pseudo data set, and merges the pseudo data set with the original data set into an enhanced data set; inputs the enhanced data set into an artificial neural network for training, and outputs the classification results of the deception state. The advantages of generative adversarial networks are used to enhance the ability of the network model to detect the deception state of the target receiver, and the selected characteristic parameters are mainly for the parameters of the baseband signal processing stage, achieving a more efficient and practical anti-interference technical effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a structural block diagram of the first embodiment of the single-antenna GNSS spoofing interference detection system based on CGAN in this application.

[0041] Figure 2 This is a framework diagram of a specific implementation plan of a single-antenna GNSS spoofing interference detection system based on CGAN in this application.

[0042] Figure 3 This is the CGAN network structure diagram of the single-antenna GNSS spoofing interference detection system based on CGAN in this application.

[0043] Figure 4 This is the artificial neural network structure diagram of the single-antenna GNSS deception interference detection system based on CGAN in this application.

[0044] Figure 5 It is a flow chart of the first embodiment of the single-antenna GNSS spoofing interference detection method based on CGAN in this application. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below through the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0046] Reference Figure 1 , Figure 1 This is a structural block diagram of the first embodiment of the single-antenna GNSS spoofing interference detection system based on CGAN in this application.

[0047] like Figure 1 As shown, the single-antenna GNSS spoofing interference detection system based on CGAN proposed in the embodiment of the present application includes:

[0048] The front-end processing module 10 is used to receive the GNSS navigation signal through the antenna and down-convert the navigation signal into a digital intermediate frequency signal;

[0049] A baseband signal processing module 20, connected to the front-end processing module, is used to capture and track the digital intermediate frequency signal, generate a local carrier and a pseudo code signal, and extract the leading, immediate and lagging correlation function values ​​of the IQ dual-path correlator;

[0050] A data set processing module 30, connected to the baseband signal processing module, is used to extract characteristic parameters from the correlation function value, and perform standardization processing on the characteristic parameters to generate a training data set;

[0051] The CGAN-ANN network module 40 is connected to the data set processing module, and includes a conditional generative adversarial network and an artificial neural network, wherein the conditional generative adversarial network is used to enhance the training data set to generate a pseudo data set, and the artificial neural network is used to perform deception status detection based on the enhanced data set and output a judgment result of deception or non-deception.

[0052] It is understood that the terms involved in this embodiment are explained as follows:

[0053] PyTorch: is an open source Python machine learning library for building and training neural networks. It is widely used in various machine learning tasks.

[0054] Machine learning: A branch of artificial intelligence that allows computers to learn from data without being explicitly programmed. Machine learning algorithms use a variety of techniques, including neural networks, decision trees, and support vector machines, to identify patterns in data and make predictions.

[0055] Deep learning: The concept of deep learning originates from the study of artificial neural networks. A multi-layer perceptron with multiple hidden layers is a deep learning structure. Deep learning combines low-level features to form more abstract high-level representations of attribute categories or features to discover distributed feature representations of data.

[0056] Generative Adversarial Network: It is a relatively large family in deep learning. Its main function is to generate data such as images, music or text. Its main idea is to train the generator and the discriminator through continuous confrontation until the discriminator cannot distinguish whether the received samples are from real samples or generated samples.

[0057] It should be noted that, in view of the many and complex problems in the selection of specific feature sets in existing artificial intelligence-based inspection technologies, a simpler feature set combination is proposed. This feature set combination can be obtained without changing the receiver hardware system. It only combines the relevant features extracted from the RF receiving end and tracking module of the GNSS receiver, and the parameter feature acquisition is easy.

[0058] In order to solve the problem of weak generalization ability of the model network, this embodiment proposes a spoofing signal detection system based on CGAN to automatically enhance the data set. The detector can also complete detection with high recognition rate for complex and unseen attack scenarios. The generalization ability of the GNSS spoofing detection model is fully improved. The GNSS spoofing detection system consists of four modules: front-end processing module, baseband signal processing module, data set processing module, and CGAN-ANN network model, which together constitute the GNSS spoofing detection system, providing a new solution for how to deal with severe GNSS security issues.

[0059] In the specific implementation, Figure 2As shown, the execution body of this embodiment first receives the navigation signals of all visible satellites through the antenna by the front-end processing module, and then the navigation signals are down-converted into digital intermediate frequency signals by the RF front-end processing unit; then the baseband signal processing module copies the local carrier and local pseudo code signal consistent with the received satellite signal by processing the digital intermediate frequency signal output by the RF front-end, so as to realize the capture and tracking of the GNSS signal, wherein the six correlation function values ​​of the lead, immediate and delayed of the I and Q paths are calculated respectively through the IQ dual-path correlator; then the navigation deception signals in all scenarios are processed by the data set processing module, and the selected characteristic parameters are extracted, and the data are standardized after the signal characteristic parameters are merged to complete the processing of the data set. Finally, the data set is input into the CGAN-ANN network model for training and deception detection and identification. In short, the CGAN network first outputs an enhanced data set with deception or non-deception labels, which will enhance the generalization ability of the network; then the enhanced data set is merged with the original data set and input into the ANN deep learning network model for training. The sigmoid activation function is used as the output function of prediction judgment. The function range is between [0,1]. When the output value is less than 0.5, it is judged as a non-deception state. When the output value is greater than 0.5, it is judged as a deception state. After overall data training and testing, the final network model parameters are output, which can effectively identify deception behaviors in various scenarios.

[0060] It should be noted that the characteristic parameters include at least one of the following: received power, which is determined by calculating the average power of the RF output signal within a preset time window; tracking loop related parameters, including phase code offset and energy loss ratio; statistical characteristics based on moving average and moving variance.

[0061] It can be understood that the CGAN includes: a generator, which receives a noise vector and a conditional label as input and generates a pseudo sample consistent with the real data distribution; a discriminator, which is used to distinguish the pseudo sample from the real sample and optimize the network parameters through adversarial training with the generator; wherein the hidden layer of the generator adopts the ReLU activation function, and the output layer adopts the Tanh activation function; the hidden layer of the discriminator adopts the LeakyReLU activation function, and the output layer adopts the Sigmoid activation function.

[0062] It should be noted that the artificial neural network includes an input layer, at least two fully connected hidden layers and an output layer, wherein: the input layer receives standardized feature parameters; the hidden layer adopts the ReLU activation function; the output layer adopts the Sigmoid activation function, and the output value is a deception probability between 0 and 1. When the output value is greater than 0.5, it is judged to be a deception state, otherwise it is a non-deception state.

[0063] In a specific implementation, the data set processing module is further used to label the multi-scenario deception signals in the TEXBAT data set, where the normal state is labeled as 0 and the deception state is labeled as 1.

[0064] In the specific implementation, the spoofing signal dataset used in this embodiment is TEXBAT, which is a high-fidelity digital real-time GPSL1 C / A code spoofing dataset with multiple sets of different spoofing scenarios provided by the Radio Navigation Laboratory of the University of Texas at Austin. The TEXBAT data contains a total of 8 different spoofing scenarios and 2 non-spoofing scenario data for comparison with the spoofing scenarios. It provides rich measured data for the research of satellite navigation spoofing jamming technology, including a large number of spoofing scenario details. Based on the different scenario GPS spoofing datasets provided by TEXBAT, a GPS spoofing jamming detection model is established.

[0065] Therefore, the working process of the front-end processing module 10 and the baseband signal processing module 20 is as follows:

[0066] First, the front-end processing module receives the navigation signals of all visible satellites through the antenna, and then the RF front-end processing unit down-converts the navigation signals into digital intermediate frequency signals, namely the TEXBAT intermediate frequency signal data set; then the baseband signal processing module processes the digital intermediate frequency signals output by the RF front-end to copy the local carrier and local pseudo code signals consistent with the received satellite signals, thereby realizing the capture and tracking of GNSS signals. The IQ dual-path correlator is used to calculate the six correlation function output values ​​of the I and Q paths, namely the advance, immediate and delayed ones.

[0067] It should be noted that the method of measuring the degree of distortion of the correlation peak of the GNSS code tracking loop is called SQM technology. A large number of studies have shown that SQM can be used to detect deceptive interference. This is different from most of the literature based on artificial intelligence, which also requires the use of other parameters such as PVT solution.

[0068] It can be understood that this embodiment extracts features from the IQ correlation value between the RF end of the GNSS receiver and the tracking loop correlator, thereby establishing a more effective and universal solution, namely, the target signal power directly received by the GNSS receiver antenna and the related feature parameters extracted by the tracking loop.

[0069] ① Receive power

[0070] For the target signal power directly received by the GNSS receiver antenna, it is assumed that Indicates RF output and receiving power By calculation arrive The average power in the interval is determined as:

[0071]

[0072] ② Tracking loop related parameters

[0073] Since there are many relevant characteristic parameters of tracking loops, and most of the parameters are essentially deformations of the phase detection parameters, these parameters can describe the state changes when deception occurs to a certain extent. Therefore, the tracking loop related parameters in the present invention are not limited to specific parameter characteristics. This article only gives examples of parameter characteristics extracted from two tracking loops. The present invention is applicable to any parameters. The best parameter combination needs to be obtained after a large number of experiments.

[0074] PCS

[0075]

[0076] in, , and Respectively The early, middle and late in-phase branch correlator outputs at the time, , and Respectively The early, middle and late orthogonal branch correlator outputs at the time, is the correlator interval.

[0077] ELP

[0078] -

[0079] ③ MA and MV

[0080] Since the original values ​​of different parameter indicators fluctuate greatly during the interaction stage, in order to better utilize them for deception detection, considering that the variance is specifically used to show the dispersion of a set of data, it can capture the changes in the correlator output during the deception attack. In addition, a method based on moving average (MA) is introduced. MA can capture changes and features over several milliseconds. Therefore, the MV and MA of the eigenvalues ​​can be intuitively regarded as a good metric supplement for detecting the occurrence of deception behavior. By bringing the original values ​​into the MV and MA algorithms respectively, two sets of data sets of moving variance and moving average corresponding to a set of original values ​​can be obtained.

[0081] MV

[0082]

[0083] in It is the first The value of the SQM sample, is the length of the MV window, is the sliding interval, is the total number of windows.

[0084] MA

[0085]

[0086] After capturing and tracking the corresponding satellite signals of the nine scenes in turn, the output values ​​of the six correlators in each scene are obtained. Assuming that five tracking loop parameters are extracted, the output values ​​of the six correlators in each scene are obtained according to the above feature extraction model. The 11 eigenvalues ​​of are used as the data of each sample.

[0087] As a supervised learning deception detection method, since the deception time in each TEXBAT scenario is roughly determined, the deception state at each moment can be labeled, with 0 in the normal state and 1 in the deception state. Finally, the feature sets obtained in each scenario are integrated, and a total of 180,000 sample data are obtained, and the dimension size of each sample data is , based on the obtained training data set, the deception detection model training experiment can be carried out.

[0088] It can be understood that the CGAN-ANN network module 40 is composed of a conditional generative adversarial network CGAN and an artificial neural network ANN. The network structure of the CGAN network is as follows: Figure 3 As shown, during the pseudo data generation process, 100 samples that follow a normal distribution are randomly generated. The noise is directly connected with the random label "0" or "1" and input into the generator in series. After passing through the generator, the pseudo sample is output. The discriminator is trained by the real label and data. The discriminator judges the forged data and outputs the loss function to alternately train the generator and the discriminator. Through continuous adversarial training, the generator and the discriminator reach a balance state. Among them, the generator adopts a fully connected structure, the activation function of the hidden layer adopts ReLU, and the activation function from the hidden layer to the output layer adopts the Tanh function; the discriminator also adopts a fully connected structure, the activation function of the hidden layer adopts the LeakyReLU function, and the activation function from the hidden layer to the output layer adopts the Sigmoid function. Save the final trained CGAN network, generate a series of pseudo data sets and merge them with the real data set to generate an enhanced data set, which are input into the deception recognition network together to judge the deception state.

[0089] In the specific implementation, Figure 4As shown in the artificial neural network structure diagram, the deception recognition network adopts the ANN network model, in which the input layer is the number of parameter eigenvalues, and the activation function from the input layer to the hidden layer adopts the ReLU function; a total of two fully connected layers are set in the hidden layer, but not limited to two fully connected layers. The specific number of layers can be designed according to actual conditions, and the activation function RELU is used between the hidden layers; the activation function from the hidden layer to the output layer adopts the sigmoid function, and the output That is, the deception status value, 0 represents the non-deception state, and 1 represents the deception state.

[0090] It should be noted that in this embodiment, the following are achieved:

[0091] Simplify characteristic parameters and reduce system complexity: By extracting the received power, tracking loop related parameters (such as PCS, ELP) and statistical features (MA, MV) from the baseband signal processing module, redundant message solution parameters such as clock error and pseudorange are discarded, and the feature set is significantly simplified. There is no need to modify the receiver hardware, which reduces the implementation cost and system complexity.

[0092] Data enhancement improves generalization ability: The conditional generative adversarial network (CGAN) is used to generate pseudo samples that conform to the real data distribution, effectively solving the problem of imbalanced training data. It is combined with the ANN model for multi-scenario training to significantly improve the model's detection robustness for unknown deception scenarios. Verification results show that the accuracy of unknown scene detection is increased by nearly 10%.

[0093] Optimization of detection efficiency and reliability: A fully connected ANN network structure is adopted, combined with the Sigmoid function to output the deception probability value (0~1), and high-precision classification is achieved through threshold judgment (>0.5 is a deception state). The detection response speed is fast and the real-time performance is strong. In complex electromagnetic environments, it can still maintain a test set detection accuracy of more than 99%.

[0094] Universality and verification guarantee: The leave-one-out verification module is used to evaluate the generalization capability of unknown scenarios, ensuring that the system has stable performance in a variety of deception attacks. It provides an efficient, economical and easy-to-deploy anti-interference solution for single-antenna GNSS receivers, which has broad application prospects.

[0095] It can be understood that the specific technical problems solved by this embodiment are:

[0096] 1. There are many types of feature parameters. In the process of researching feature indicators established using the output values ​​of related loops, there are many types of combinations. It is necessary to cover features that can describe the trend of related loop changes and reduce redundant features as much as possible. In addition, for each deception scenario, different features have different sensitivities to loop changes in different scenarios. Therefore, it is necessary to design appropriate data processing strategies to select the best feature combination.

[0097] 2. Design reasonable generators and discriminators to make the CGAN network converge more stably. CGAN introduces conditional variables based on traditional GAN, so that the model not only generates samples similar to real data, but also combines the deception / non-deception states corresponding to the features to meet specific conditions. This involves how to design the generator and discriminator networks so that the generated data has both high fidelity and good diversity under conditional constraints. In addition, the training process of the generative adversarial network is unstable, which may lead to mode collapse or training failure. In CGAN, with the addition of deception state conditional variables, this instability may be further exacerbated. How to balance the competitive relationship between the generator and the discriminator during training to ensure that the model can converge stably is an important scientific issue.

[0098] The front-end processing module and the baseband signal processing module are general. Therefore, among the four major modules of this system, the main focus is on the data set processing module and the CGAN-ANN network module.

[0099] In the specific implementation, in order to verify the generalization ability of the selected feature parameters and network model, a deception detection experiment in an unknown scenario is conducted, that is, deception detection is conducted when the deception interference source of the detection object has unknown working parameters. The experimental method adopted is the leave-one-out experiment, and the specific classification method of the data set is: select the data set samples in any scenario as the verification set, that is, for the scene data set where both the training set and the test set are unknown, and then randomly divide the data set of the remaining scenes into the training set and the test set according to the ratio of 70% and 30%. Therefore, there are 112,000 training set samples, 48,000 test set samples and 20,000 verification set samples in the end. The network model is trained by the training set and the test set, and then the verification set is used to verify the model's deception detection ability in unknown scenarios. To a certain extent, it can be used as an evaluation indicator of the generalization ability of the selected feature parameters and network models.

[0100] Table 1 compares the deception detection effects of the classic ANN model and the algorithm CGAN-ANN proposed in this paper. From the data analysis, it can be seen that the deception detection effect in the unknown scene of ds7 is the worst, but even in the ANN verification set, the recognition rate can still reach 84.30%, and the overall detection effect is good. Comparing the verification set accuracy of the ANN model and the CGAN-ANN model, it can be seen that the detection rate of the CGAN-ANN model for deception signals is generally better than that of the ANN model. The numbers on both sides of the " / " in the table represent the recognition rate of the test set and the recognition rate of the verification set, respectively.

[0101] Table 1 Comparison of ANN and CGAN-ANN network deception detection recognition rates in unknown scenarios

[0102]

[0103] This embodiment receives GNSS navigation signals through antennas and down-converts the signals into digital intermediate frequency signals; captures and tracks digital intermediate frequency signals, extracts the leading, immediate and lagging correlation function values ​​of the IQ dual-path correlator; extracts characteristic parameters from the correlation function values ​​and performs standardization processing to generate a training data set; uses CGAN to enhance the training data set, generates a pseudo data set, and merges the pseudo data set with the original data set into an enhanced data set; inputs the enhanced data set into an artificial neural network for training, and outputs the classification results of the deception state. The advantages of the generative adversarial network are used to enhance the network model's ability to detect the deception state of the target receiver, and the selected characteristic parameters are mainly for the parameters of the baseband signal processing stage, achieving a more efficient and practical anti-interference technical effect.

[0104] In addition, an embodiment of the present application also proposes a computer-readable storage medium, on which a program for single-antenna GNSS spoofing interference detection based on CGAN is stored. When the program for single-antenna GNSS spoofing interference detection based on CGAN is executed by a processor, the steps of the method for single-antenna GNSS spoofing interference detection based on CGAN as described above are implemented.

[0105] It should be understood that the above is only an example and does not constitute any limitation on the technical solution of the present application. In specific applications, technicians in this field can make settings as needed, and the present application does not impose any limitation on this.

[0106] like Figure 5 As shown, Figure 5 It is a flow chart of the first embodiment of the single-antenna GNSS spoofing interference detection method based on CGAN in this application.

[0107] The single-antenna GNSS spoofing interference detection method based on CGAN proposed in the embodiment of the present application includes:

[0108] S1: Receive GNSS navigation signals through an antenna and down-convert the signals into digital intermediate frequency signals.

[0109] S2: Capture and track the digital intermediate frequency signal, and extract the leading, immediate and delayed correlation function values ​​of the IQ dual-path correlator.

[0110] S3: Extract characteristic parameters from the correlation function values ​​and perform standardization processing to generate a training data set.

[0111] S4: Use CGAN to enhance the training dataset to generate a pseudo dataset, and merge the pseudo dataset with the original dataset into an enhanced dataset.

[0112] In a specific implementation, the generator of the CGAN receives a noise vector and a conditional label to generate a pseudo sample, and the discriminator optimizes the parameters of the generator and the discriminator through adversarial training until the distribution of the generated samples is consistent with that of the real samples.

[0113] S5: Input the enhanced data set into ANN for training, and output the classification result of the deception state.

[0114] It should be noted that the Sigmoid function is used as the output layer activation function of the ANN. When the output value is greater than 0.5, it is judged to be a deception state, otherwise it is a non-deception state.

[0115] This embodiment receives GNSS navigation signals through antennas and down-converts the signals into digital intermediate frequency signals; captures and tracks digital intermediate frequency signals, extracts the leading, immediate and lagging correlation function values ​​of the IQ dual-path correlator; extracts characteristic parameters from the correlation function values ​​and performs standardization processing to generate a training data set; uses CGAN to enhance the training data set, generates a pseudo data set, and merges the pseudo data set with the original data set into an enhanced data set; inputs the enhanced data set into an artificial neural network for training, and outputs the classification results of the deception state. The advantages of the generative adversarial network are used to enhance the network model's ability to detect the deception state of the target receiver, and the selected characteristic parameters are mainly for the parameters of the baseband signal processing stage, achieving a more efficient and practical anti-interference technical effect.

[0116] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present application. In practical applications, technicians in this field can select part or all of it according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0117] In addition, for technical details not described in detail in this embodiment, please refer to the single-antenna GNSS spoofing interference detection method based on CGAN provided in any embodiment of the present application, which will not be repeated here.

[0118] In addition, it should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0119] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0120] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory (ROM) / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of each embodiment of the present application. The above is only a preferred embodiment of the present application, and does not limit the patent scope of the present application. All equivalent structures or equivalent process changes made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are similarly included in the patent protection scope of the present application.

Claims

1. A single-antenna GNSS spoofing jamming detection system based on CGAN, characterized in that: include: A front-end processing module, used for receiving GNSS navigation signals through an antenna and down-converting the navigation signals into digital intermediate frequency signals; A baseband signal processing module, connected to the front-end processing module, is used to capture and track the digital intermediate frequency signal, generate a local carrier and a pseudo code signal, and extract the leading, immediate and lagging correlation function values ​​of the IQ dual-path correlator; A data set processing module, connected to the baseband signal processing module, for extracting characteristic parameters from the correlation function values, and performing standardization processing on the characteristic parameters to generate a training data set; The CGAN-ANN network module is connected to the data set processing module and includes a conditional generative adversarial network and an artificial neural network, wherein the conditional generative adversarial network is used to enhance the training data set to generate a pseudo data set, and the artificial neural network is used to perform deception state detection based on the enhanced data set and output a judgment result of deception or non-deception.

2. The system according to claim 1, characterized in that The characteristic parameters include at least one of the following: The received power is determined by calculating the average power of the RF output signal within a preset time window; Tracking loop related parameters, including phase code offset and energy loss ratio; Statistical characteristics based on moving average and moving variance.

3. The system according to claim 1, characterized in that The CGAN includes: The generator receives the noise vector and the conditional label as input and generates pseudo samples that are consistent with the real data distribution; A discriminator, used to distinguish the pseudo samples from real samples, and optimize network parameters through adversarial training with the generator; Among them, the hidden layer of the generator adopts the ReLU activation function, and the output layer adopts the Tanh activation function; the hidden layer of the discriminator adopts the LeakyReLU activation function, and the output layer adopts the Sigmoid activation function.

4. The system according to claim 1, characterized in that The artificial neural network comprises an input layer, at least two fully connected hidden layers and an output layer, wherein: The input layer receives the standardized feature parameters; The hidden layer adopts ReLU activation function; The output layer adopts a Sigmoid activation function, and the output value is a deception probability between 0 and 1. When the output value is greater than 0.5, it is determined to be a deception state, otherwise it is a non-deception state.

5. The system according to claim 1, characterized in that The data set processing module is further used to label the multi-scenario deception signals in the TEXBAT data set, where the normal state is labeled as 0 and the deception state is labeled as 1.

6. The system according to claim 1, characterized in that The system further includes a verification module for verifying the deception detection performance of unknown scenarios by using a leave-one-out method, specifically including: The data of any scenario is selected as the validation set, and the remaining scenario data is divided into training set and test set in proportion; The CGAN-ANN network module is trained through a training set and a test set, and the generalization ability of the model is evaluated through a validation set.

7. The system according to claim 1, characterized in that The baseband signal processing module outputs six kinds of correlation function values ​​through an IQ dual-path correlator, including leading, immediate and lagging correlation values ​​of the I path and the Q path.

8. A single-antenna GNSS spoofing interference detection method based on CGAN, characterized in that: include: S1: Receive GNSS navigation signals through an antenna and down-convert the signals into digital intermediate frequency signals; S2: Capture and track the digital intermediate frequency signal, and extract the leading, immediate and delayed correlation function values ​​of the IQ dual-path correlator; S3: extracting characteristic parameters from the correlation function values ​​and performing standardization processing to generate a training data set; S4: using CGAN to enhance the training dataset to generate a pseudo dataset, and merging the pseudo dataset with the original dataset into an enhanced dataset; S5: Input the enhanced data set into an artificial neural network for training, and output a classification result of the deception state.

9. The method according to claim 8, characterized in that In step S4, the generator of the CGAN receives the noise vector and the conditional label to generate a pseudo sample, and the discriminator optimizes the parameters of the generator and the discriminator through adversarial training until the distribution of the generated samples is consistent with that of the real samples.

10. The method according to claim 8, characterized in that In step S5, the Sigmoid function is used as the output layer activation function of the ANN, and when the output value is greater than 0.5, it is determined to be a deception state, otherwise it is a non-deception state.

Images