Anti-spoofing and anti-interference satellite positioning security protection method
By combining vector tracking, adaptive filtering, and a lightweight machine learning model in a satellite positioning system, an adaptive protection closed loop is constructed, which solves the problems of deception and interference in complex electromagnetic environments, and achieves efficient positioning recovery and anti-interference capabilities, making it suitable for UAV platforms.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SI CHUAN KE RUI RUAN JIAN YOU XIAN ZE REN GONG SI
- Filing Date
- 2026-03-31
- Publication Date
- 2026-07-21
Smart Images
Figure CN121956047B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of satellite positioning security protection, specifically to a satellite positioning security protection method that resists deception and interference. Background Technology
[0002] The reliability of navigation information and mission reliability for unmanned aerial vehicles (UAVs) in highly contested and complex electromagnetic environments are crucial. With the rapid development of the low-altitude economy, UAVs are increasingly used in critical tasks such as material transport and emergency rescue. Satellite navigation systems (such as GPS, BeiDou, GLONASS, and Galileo) provide indispensable positioning, navigation, and timing services. However, in complex low-altitude operating environments, especially in urban canyons, mountainous areas, forest areas, or near near-ground electromagnetic interference sources, satellite signals are not only susceptible to physical obstruction and multipath effects, but also face increasingly severe threats of malicious human interference. These threats can be mainly divided into two categories:
[0003] Suppression jamming: By transmitting high-power broadband or narrowband noise signals, the real satellite navigation signals are overwhelmed, causing the receiver to be unable to capture or track the signals, resulting in a complete interruption of positioning services.
[0004] Deceptive jamming: By transmitting forged signals that are highly similar to real satellite signal parameters (code phase, carrier frequency, navigation message), the receiver is induced to output incorrect position, velocity, or time information. This type of attack is highly covert and extremely dangerous, potentially causing drones to veer off course, crash, or even be hijacked, posing a serious threat to mission safety and public safety.
[0005] Currently, protection technologies against satellite navigation interference are mostly focused on a single level or employ isolated strategies, for example:
[0006] At the signal processing level, spatial filtering (such as null antennas) or frequency filtering is used to suppress broadband interference in specific directions.
[0007] At the information processing level, receiver autonomous integrity monitoring or multi-receiver consistency verification is used to detect large errors.
[0008] However, existing technologies have significant limitations:
[0009] Traditional detection methods based on signal power or simple consistency have a high false negative rate for intelligent and dynamically changing deceptive signals.
[0010] Under strong suppression interference, the receiver front end may become saturated, causing all subsequent processing stages to fail.
[0011] The lack of a closed-loop collaborative protection mechanism from signal preprocessing to information decision-making means that the protection measures at each stage have failed to work together effectively.
[0012] Many highly complex algorithms are difficult to run in real time on embedded platforms such as unmanned helicopters, which have strict limitations on size, weight, and power consumption. Summary of the Invention
[0013] The purpose of this invention is to overcome the shortcomings of the prior art and provide a satellite positioning security protection method that is resistant to deception and interference, thereby improving the anti-interference capability and the accuracy of deception detection.
[0014] The present invention achieves the above objectives by adopting the following technical solution: the present invention provides a satellite positioning security protection method against deception and interference, comprising:
[0015] S1, Anti-interference baseband signal depth tracking processing;
[0016] S2. In the baseband signal and tracking loop processed in step S1, multi-level feature vectors are extracted in real time. The feature vectors include: signal power layer features, correlation function layer features, carrier measurement layer features, navigation information layer features, and signal quality layer features.
[0017] S3. Offline training phase: Construct training dataset and perform supervised training on the selected lightweight machine learning model.
[0018] A training dataset is constructed, which contains multiple samples. Each sample consists of five feature vectors extracted from real signals, various known spoofing signals, and suppressed interference signals, and is labeled with the corresponding signal category label.
[0019] The selected lightweight machine learning model is trained in a supervised manner using the training dataset, so that the model learns the mapping relationship between different feature combinations and signal categories;
[0020] The five types of feature vectors include: signal power layer features, correlation function layer features, carrier measurement layer features, navigation information layer features, and signal quality layer features;
[0021] The known spoofing signals include forwarding spoofing signals, generating spoofing signals, and hybrid spoofing signals; the suppression jamming signals include high-power suppression jamming signals; the signal category labels include real, forwarding spoofing, generating spoofing, and suppression jamming.
[0022] After training, the lightweight machine learning model outputs the probability distribution of the current signal belonging to each category based on the input feature vector; based on this probability distribution, a comprehensive deception threat index is calculated through a preset fusion rule, which is positively correlated with the severity of the threat.
[0023] S4. In the online detection phase, the feature vectors extracted in real time are input into the trained lightweight machine learning model for inference, and a comprehensive deception threat index is output.
[0024] The feature vectors extracted in real time are input into a pre-trained lightweight machine learning model for inference to obtain the deception threat index;
[0025] The deception threat index is compared with a dynamic threshold, which is adjusted in real time based on one or more factors, including the current satellite geometry, signal environment, and historical false alarm rate.
[0026] If the deception threat index exceeds the dynamic threshold, it is determined that there is a deception or interference threat.
[0027] Based on the category probabilities or feature activation patterns output by the lightweight machine learning model, specific threat types are identified.
[0028] The specific identification types include: if the signal power layer features and the correlation function layer features are both significantly abnormal, it is determined to be a suppression-type deception; if the abnormality is concentrated in the correlation function layer and the signal quality layer features, it is determined to be a generative deception.
[0029] S5. If step S4 determines that there is a deception or interference threat, then perform parameter estimation and active suppression or reconstruction of the deception or interference signal.
[0030] S6. Combining the tracking results of step S1, the threat assessment results of step S4, and the suppression or reconstruction results of S5, execute the final security decision logic.
[0031] Furthermore, step S1 specifically includes:
[0032] The system receives satellite radio frequency signals and performs down-conversion and analog-to-digital conversion to obtain digital intermediate frequency signals. An improved vector delay-locked loop and vector frequency-locked loop joint tracking architecture is adopted to replace the traditional scalar tracking loop. In this architecture, the code phase error and carrier frequency error of all channels are input into a centralized Kalman filter for state estimation and prediction. The spatial geometric constraints between satellite signals are used to improve the overall tracking robustness. At the same time, an adaptive interference suppression filter is embedded in the baseband processing link to estimate the spectral characteristics of interference signals in real time and perform notch filtering in the frequency domain.
[0033] Furthermore, in step S2, the signal power layer characteristics include the short-term fluctuation of the carrier-to-noise ratio and the automatic gain control level. If the signal power layer characteristics are abnormal, it is associated with the abnormal rise of signal power in the early stage of suppression interference or forwarding spoofing attack.
[0034] The correlation function layer features include the symmetry deviation of the output values of the early and late correlators and the spatial consistency index of the peak values of the multiple correlators. If the correlation function layer features are abnormal, that is, the correlation peaks are asymmetrical and the distribution of the peak values of the multiple correlators does not match the expectations, then it is more sensitive to generative spoofing or precise forwarding spoofing.
[0035] The carrier measurement layer features include carrier phase continuity measurement, Doppler frequency shift rate, and short-term calculation results from the inertial measurement unit. If the carrier measurement layer features are abnormal, i.e., the carrier phase jumps or the Doppler and inertial measurement unit calculation results are inconsistent, the abnormal carrier measurement layer features are used to detect generative spoofing or precise forwarding spoofing.
[0036] The navigation information layer features include the deviation between the calculated position or velocity and the historical trajectory, and the residuals between the calculated position and the predicted values of the aircraft dynamics model. Anomalies in the navigation information layer features include position jumps that do not conform to the laws of dynamics, and velocity contradictions with the historical trajectory.
[0037] This indicates that the navigation message has been tampered with or that there is a dynamic contradiction in the position or velocity information;
[0038] Signal quality layer features include chip waveform distortion and carrier phase noise level. These features reflect the distortion introduced during signal propagation or generation. Anomalies in signal quality layer features are used to identify non-standard signal sources.
[0039] Furthermore, in step S5, parameter estimation: for the identified spoofing or signal, key parameters are quickly estimated. The key parameters include interference type, center frequency, bandwidth, and signal strength. Interference type includes continuous wave, frequency sweep, and spoofing.
[0040] Active suppression: For suppression interference, an adaptive cancellation algorithm is used in the digital domain. Based on the estimated parameters, an inverted copy of the interference signal is generated and added to the original intermediate frequency signal, thereby achieving effective cancellation of interference in the digital link.
[0041] Signal reconstruction: For spoofing signals, if some satellite signals are confirmed to be reliable, the pseudorange and carrier phase observations of the satellite channels affected by the spoofing signals are reconstructed by using the reliable signals and the short-time high-precision inertial calculation results from the inertial measurement unit through a tightly coupled filtering algorithm, thereby restoring a usable and reliable navigation solution.
[0042] Furthermore, in step S6, the final security decision logic is as follows:
[0043] Generate a safety weighting factor to weight observations from different satellite signals or different frequencies;
[0044] In the localization solution filter, weighted observations are used for solution;
[0045] Output the final hardened positioning, speed, and time information, and also output the system safety status indicator.
[0046] The beneficial effects of this invention are as follows:
[0047] This invention combines vector tracking with adaptive frequency domain filtering, enabling the receiver to maintain stable signal tracking and positioning output under stronger interference suppression environments than traditional methods, thus significantly improving the system's survivability and anti-interference capability.
[0048] This invention comprehensively utilizes multi-dimensional features from the physical layer to the application layer, and combines them with machine learning models to detect various deception signals, including advanced generative deception, with a high probability. It has low average detection latency and low false alarm rate, which is superior to traditional single-feature detection methods.
[0049] This invention can not only detect threats, but also partially restore the availability and accuracy of location services when interference exists through proactive cancellation or information reconstruction technology, thus achieving a leap from "passive alarm" to "proactive defense and recovery".
[0050] This invention tightly couples signal hardening, threat detection, and active suppression to form a complete adaptive security protection closed loop, which improves the overall resilience and intelligence of the system in the face of complex and dynamic interference.
[0051] This invention meets the real-time requirements of embedded systems. Through algorithm optimization and lightweight model design, it can run in real time on typical embedded processors with low processing latency and controllable power consumption, making it fully applicable to UAV platforms. Attached Figure Description
[0052] Figure 1 This is a flowchart of a satellite positioning security protection method against deception and interference provided by an embodiment of the present invention. Detailed Implementation
[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.
[0054] This invention provides a satellite positioning security protection method that resists deception and interference, such as... Figure 1 As shown, it includes:
[0055] S1, Anti-interference baseband signal depth tracking processing;
[0056] The system receives satellite radio frequency signals and performs down-conversion and analog-to-digital conversion to obtain digital intermediate frequency (IF) signals. An improved vector delay-locked loop (DLL) and vector frequency-locked loop (VLL) joint tracking architecture replaces the traditional scalar tracking loop. In this architecture, the code phase error and carrier frequency error of all channels are jointly input into a centralized Kalman filter for state estimation and prediction, utilizing the spatial geometric constraints between satellite signals to improve overall tracking robustness. Simultaneously, an adaptive interference suppression filter is embedded in the baseband processing link to estimate the spectral characteristics of interference signals in real time and perform notch filtering in the frequency domain to mitigate the impact of interference suppression on the correlator.
[0057] S2. Intelligent detection and recognition of multi-dimensional deception signals;
[0058] In the baseband signal and tracking loop processed in step S1, a multi-level feature vector is extracted in real time. The feature vector includes:
[0059] Signal power layer characteristics, including short-term fluctuations in carrier-to-noise ratio and automatic gain control level, are associated with abnormal signal power increases in the early stages of suppression interference or forwarding spoofing attacks.
[0060] The correlation function layer features include the symmetry deviation of the output values of the early and late correlators and the spatial consistency index of the peak values of the multiple correlators. If the correlation function layer features are abnormal, that is, the correlation peaks are asymmetrical and the distribution of the peak values of the multiple correlators does not match the expectations, then it is more sensitive to generative spoofing or precise forwarding spoofing.
[0061] The carrier measurement layer characteristics include carrier phase continuity measures, Doppler frequency shift rate, and short-term extrapolation results from the inertial measurement unit (IMU). If the carrier measurement layer characteristics are abnormal, such as carrier phase jumps or inconsistencies between Doppler and IMU extrapolation results, then...
[0062] The carrier measurement layer feature anomalies are used to detect generative spoofing or precision forwarding spoofing;
[0063] The navigation information layer features include the deviation between the calculated position or velocity and the historical trajectory, and the residuals between the calculated position and the predicted values of the aircraft dynamics model. Anomalies in the navigation information layer features include position jumps that do not conform to the laws of dynamics, and velocity contradictions with the historical trajectory.
[0064] This indicates that the navigation message has been tampered with or that there is a dynamic contradiction in the position or velocity information;
[0065] Signal quality layer features include chip waveform distortion and carrier phase noise level. These features reflect the distortion introduced during signal propagation or generation. Anomalies in signal quality layer features are used to identify non-standard signal sources.
[0066] S3. Offline training phase: Construct training dataset and perform supervised training on the selected lightweight machine learning model.
[0067] Construct a training dataset containing multiple samples. Each sample consists of five feature vectors extracted from real signals, various known spoofing signals, and suppressed interference signals, and is labeled with the corresponding signal category label.
[0068] Supervised training is performed on a selected lightweight machine learning model using a training dataset, enabling the model to learn the mapping relationship between different feature combinations and signal categories;
[0069] The five types of feature vectors include: signal power layer features, correlation function layer features, carrier measurement layer features, navigation information layer features, and signal quality layer features;
[0070] The known spoofing signals include forwarding spoofing signals, generating spoofing signals, and hybrid spoofing signals; the suppression and interference signals include high-power suppression and interference signals; the signal category labels include real, forwarding spoofing, generating spoofing, and suppression and interference.
[0071] After training, the model outputs the probability distribution of the current signal belonging to each category based on the input feature vector; based on this probability distribution, a comprehensive deception threat index is calculated through a preset fusion rule, which is positively correlated with the severity of the threat.
[0072] S4. In the online detection phase, the feature vectors extracted in real time are input into the trained model for inference, and a comprehensive deception threat index is output.
[0073] The feature vectors extracted in real time are input into a pre-trained machine learning model for inference to obtain the deception threat index.
[0074] The deception threat index is compared with a dynamic threshold, which is adjusted in real time based on one or more factors, including the current satellite geometry, signal environment, and historical false alarm rate.
[0075] If the deception threat index exceeds the dynamic threshold, it is determined that there is a deception or interference threat.
[0076] Based on the probability of each category or the activation pattern of features output by the model, the specific threat type is identified;
[0077] The specific types of identification include: if the signal power layer features and the correlation function layer features are both significantly abnormal, it is determined to be a suppression-type spoofing; if the abnormalities are concentrated in the correlation function layer and the signal quality layer features, it is determined to be a generative spoofing.
[0078] S5. If step S4 determines that there is a deception or interference threat, then perform parameter estimation and active suppression or reconstruction of the deception or interference signal.
[0079] Parameter estimation: For the identified spoofing or signal, quickly estimate key parameters, including interference type, center frequency, bandwidth, and signal strength. Interference types include continuous wave, frequency sweep, and spoofing.
[0080] Active suppression: For suppression interference, an adaptive cancellation algorithm is used in the digital domain. Based on the estimated parameters, an inverted copy of the interference signal is generated and added to the original intermediate frequency signal, thereby achieving effective cancellation of interference in the digital link.
[0081] Signal reconstruction: For spoofing signals, if some satellite signals are confirmed to be reliable, the pseudorange and carrier phase observations of the satellite channels affected by the spoofing signals are reconstructed by using the reliable signals and the short-time high-precision inertial calculation results from the inertial measurement unit through a tightly coupled filtering algorithm, thereby restoring a usable and reliable navigation solution.
[0082] S6. Combining the tracking results of step S1, the threat assessment results of step S4, and the suppression or reconstruction results of S5, execute the final security decision logic.
[0083] The final security decision logic is as follows:
[0084] Generate a safety weighting factor to weight observations from different satellite signals or different frequencies;
[0085] In the localization solution filter, weighted observations are used for solution;
[0086] Output the final hardened positioning, speed, and time information, and also output the system safety status indicator.
[0087] The implementation based on the software-defined radio platform will be described in detail below.
[0088] Hardware platform: A software-defined radio platform (such as USRP) with an integrated high-speed ADC (Analog-to-Digital Converter) is used as the RF front end and connected to an embedded industrial computer.
[0089] Run the GNSS-SDR open-source framework (C++ source code) on an industrial control computer and perform secondary development to implement the vector tracking loop and adaptive filter in step S1.
[0090] A lightweight gradient boosting decision tree model was trained using Python's Scikit-learn library for deception detection in step S2. The model was then serialized and exported, and model inference was implemented using C++ code and integrated into the processing flow.
[0091] Implement the adaptive cancellation algorithm (using the NLMS algorithm) for step S5 and the tightly coupled reconstruction algorithm based on extended Kalman filtering in C++.
[0092] Design a decision scheduling manager to coordinate the pipeline execution and data interaction of steps S1 to S6.
[0093] Workflow:
[0094] The radio frequency signal is acquired by the USRP and transmitted to the industrial control computer, and the following steps are executed in sequence: S1 (hardening tracking) -> S2 (feature extraction and threat detection) -> S3 (offline training) -> S4 (online detection) -> if necessary, S5 (suppression / reconstruction) -> S6 (security weighting and calculation).
[0095] Finally, the hardened PVT information (location, speed, time) and security status are output through the Ethernet port.
[0096] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.
Claims
1. A satellite positioning security protection method that resists deception and interference, characterized in that, include: S1, Anti-interference baseband signal depth tracking processing; S2. In the baseband signal and tracking loop processed in step S1, multi-level feature vectors are extracted in real time. The feature vectors include: signal power layer features, correlation function layer features, carrier measurement layer features, navigation information layer features, and signal quality layer features. S3. Offline training phase: Construct training dataset and perform supervised training on the selected lightweight machine learning model. A training dataset is constructed, which contains multiple samples. Each sample consists of five feature vectors extracted from real signals, various known spoofing signals, and suppressed interference signals, and is labeled with the corresponding signal category label. The selected lightweight machine learning model is trained in a supervised manner using the training dataset, so that the lightweight machine learning model learns the mapping relationship between different feature combinations and signal categories. The five types of feature vectors include: signal power layer features, correlation function layer features, carrier measurement layer features, navigation information layer features, and signal quality layer features; The known spoofing signals include forwarding spoofing signals, generating spoofing signals, and hybrid spoofing signals; the suppression jamming signals include high-power suppression jamming signals; the signal category labels include real, forwarding spoofing, generating spoofing, and suppression jamming. After training, the lightweight machine learning model outputs the probability distribution of the current signal belonging to each category based on the input feature vector; based on this probability distribution, a comprehensive deception threat index is calculated through a preset fusion rule, which is positively correlated with the severity of the threat. S4. In the online detection phase, the feature vectors extracted in real time are input into the trained lightweight machine learning model for inference. The feature vectors extracted in real time are input into a pre-trained lightweight machine learning model for inference to obtain the deception threat index; The deception threat index is compared with a dynamic threshold, which is adjusted in real time based on one or more factors, including the current satellite geometry, signal environment, and historical false alarm rate. If the deception threat index exceeds the dynamic threshold, it is determined that there is a deception or interference threat. Based on the probability of each category output by the lightweight machine learning model, the specific threat type is identified; The specific identification types include: if the signal power layer features and the correlation function layer features are significantly abnormal at the same time, it is determined to be a suppression interference; if the abnormality is concentrated in the correlation function layer and the signal quality layer features, it is determined to be a generative spoofing. S5. If step S4 determines that there is a deception or interference threat, then perform parameter estimation and active suppression or reconstruction of the deception or interference signal. S6. Combining the tracking results of step S1, the threat assessment results of step S4, and the suppression or reconstruction results of S5, execute the final security decision logic.
2. The satellite positioning security protection method against deception and interference according to claim 1, characterized in that, Step S1 specifically includes: The system receives satellite radio frequency signals and performs down-conversion and analog-to-digital conversion to obtain digital intermediate frequency signals. An improved vector delay-locked loop and vector frequency-locked loop joint tracking architecture is adopted to replace the traditional scalar tracking loop. In this architecture, the code phase error and carrier frequency error of all channels are input into a centralized Kalman filter for state estimation and prediction. The spatial geometric constraints between satellite signals are used to improve the overall tracking robustness. At the same time, an adaptive interference suppression filter is embedded in the baseband processing link to estimate the spectral characteristics of interference signals in real time and perform notch filtering in the frequency domain.
3. The satellite positioning security protection method against deception and interference according to claim 1, characterized in that, In step S2, the signal power layer characteristics include the short-term fluctuation of the carrier-to-noise ratio and the automatic gain control level. If the signal power layer characteristics are abnormal, it is associated with the abnormal rise of signal power in the early stage of suppression interference or forwarding spoofing attack. The correlation function layer features include the symmetry deviation of the output values of the early and late correlators and the spatial consistency index of the peak values of the multiple correlators. If the correlation function layer features are abnormal, that is, the correlation peaks are asymmetrical and the distribution of the peak values of the multiple correlators does not match the expectations, then it is more sensitive to generative spoofing or precise forwarding spoofing. The carrier measurement layer features include carrier phase continuity measurement, Doppler frequency shift rate, and short-term calculation results from the inertial measurement unit. If the carrier measurement layer features are abnormal, i.e., the carrier phase jumps or the Doppler and inertial measurement unit calculation results are inconsistent, the abnormal carrier measurement layer features are used to detect generative spoofing or precise forwarding spoofing. The navigation information layer features include the deviation between the calculated position or velocity and the historical trajectory, and the residuals between the calculated position and the predicted values of the aircraft dynamics model. Anomalies in the navigation information layer features include position jumps that do not conform to the laws of dynamics, and velocity contradictions with the historical trajectory. This indicates that the navigation message has been tampered with or that there is a dynamic contradiction in the position or velocity information; Signal quality layer features include chip waveform distortion and carrier phase noise level. These features reflect the distortion introduced during signal propagation or generation. Anomalies in signal quality layer features are used to identify non-standard signal sources.
4. The satellite positioning security protection method against deception and interference according to claim 1, characterized in that, In step S5, parameter estimation: for the identified spoofing or interference signals, key parameters are quickly estimated. Key parameters include interference type, center frequency, bandwidth, and signal strength. Interference types include continuous wave, frequency sweep, and spoofing. Active suppression: For suppression interference, an adaptive cancellation algorithm is used in the digital domain. Based on the estimated parameters, an inverted copy of the interference signal is generated and added to the original intermediate frequency signal, thereby achieving effective cancellation of interference in the digital link. Signal reconstruction: For spoofing signals, if some satellite signals are confirmed to be reliable, the pseudorange and carrier phase observations of the satellite channels affected by the spoofing signals are reconstructed by using the reliable signals and the short-time high-precision inertial calculation results from the inertial measurement unit through a tightly coupled filtering algorithm, thereby restoring a usable and reliable navigation solution.
5. The satellite positioning security protection method against deception and interference according to claim 1, characterized in that, In step S6, the final security decision logic is as follows: Generate a safety weighting factor to weight observations from different satellite signals or different frequencies; In the localization solution filter, weighted observations are used for solution; Output the final hardened positioning, speed, and time information, and also output the system safety status indicator.