GNSS interference monitoring method and system based on ADS-B and isolation forest algorithm

Abstract

The application discloses a kind of GNSS interference monitoring method and system based on ADS-B and isolated forest algorithm, belong to GNSS interference monitoring field.The method includes: collecting and pre-processing ADS-B data to eliminate non-interference anomaly;Data is divided according to preset space-time grid and the feature reflecting GNSS quality is extracted;Based on the feature, a detection model is constructed using the isolated forest algorithm, and a forest is formed by constructing multiple isolated trees through random partitioning;The test data is input into the model to calculate the abnormal score, and whether GNSS interference occurs is determined by comparing with the threshold;Finally, the monitoring results are visualized in the form of heat map, trajectory diagram, etc.The corresponding system includes data preprocessing, feature extraction, model construction, interference detection and visualization module.The application realizes wide area monitoring using ADS-B data, efficiently processes massive data and accurately identifies interference through isolated forest algorithm, and intuitively presents the results combined with the visualization interface, significantly improves the real-time performance, accuracy and operability of civil aviation GNSS interference monitoring.

Description

Technical Field

[0001] This invention relates to the field of Global Navigation Satellite System (GNSS) interference monitoring technology, and in particular to a GNSS interference monitoring method and system based on ADS-B and the isolated forest algorithm. Background Technology

[0002] Global Navigation Satellite Systems (GNSS), with their all-weather, high-precision, and wide-coverage characteristics, have become an indispensable core navigation and surveillance infrastructure in modern civil aviation. From takeoff and departure to en-route cruise and approach and landing, GNSS continuously provides aircraft with precise position, speed, and time information, serving as a key support for ensuring flight safety and improving air traffic control efficiency. However, with the lowering of technological barriers and the widespread adoption of radio transmitting devices such as anti-drone equipment, GNSS interference incidents in civil aviation are frequent. Such interference, by transmitting electromagnetic signals of the same or adjacent frequencies, suppresses or deceives airborne GNSS receivers, causing positioning data jumps, drifts, or even failures. This can lead to serious safety hazards such as track deviations and abnormal altitudes, directly threatening flight safety.

[0003] Automatic Dependent Surveillance-Broadcast (ADS-B) technology, as the core of next-generation air traffic control surveillance, works by having aircraft autonomously acquire positioning and status information from sources such as GNSS and periodically broadcast it outwards via a specific data link. Compared to traditional radar, ADS-B offers advantages such as wider coverage, faster updates, lower cost, and richer data dimensions. Since ADS-B data originates directly from GNSS, its quality directly reflects the integrity of GNSS signals, thus enabling it to detect anomalies such as track breaks and position jumps caused by GNSS interference, making it a potential data source for GNSS interference monitoring.

[0004] Currently, the analysis of ADS-B data anomalies largely focuses on data quality assessment or simple threshold alarms, lacking mature methods specifically designed for GNSS interference as a particular cause, capable of handling massive amounts of data, and achieving efficient and automated monitoring. Existing technologies struggle to accurately and in real-time separate anomalies caused by GNSS interference from massive, high-dimensional ADS-B data streams, and the monitoring results are often not intuitive enough, hindering quick understanding and response by non-professionals. Summary of the Invention

[0005] This invention aims to overcome the shortcomings of existing technologies and provide a method and system for efficiently, accurately, and intuitively monitoring GNSS interference. Its core lies in utilizing the inherent characteristics of ADS-B data and combining the advantages of the Isolation Forest algorithm in processing massive, high-dimensional data anomaly detection to construct a complete monitoring system encompassing data preprocessing, feature analysis, anomaly detection, and result visualization.

[0006] To solve the above-mentioned technical problems, the technical solution proposed in this application is as follows:

[0007] This invention provides a GNSS interference monitoring method based on ADS-B and the isolated forest algorithm, comprising the following steps: S1. ADS-B Data Acquisition and Preprocessing: Acquire raw ADS-B data in the target airspace; preprocess the raw ADS-B data to remove non-interference abnormal data caused by equipment failure or transmission errors, and obtain the preprocessed dataset; S2. Data Spatiotemporal Division and Feature Extraction: The preprocessed dataset is divided according to a preset spatiotemporal grid; for the data in each spatiotemporal grid, a set of feature parameters related to GNSS signal quality is extracted. S3. Construction and training of the isolated forest model: Based on the feature parameter set obtained in step S2, construct the isolated forest model; S4. GNSS Interference Anomaly Detection: Input the ADS-B data to be monitored into the trained isolated forest model and calculate the anomaly score for each data sample; compare the anomaly score with a preset interference judgment threshold τ. If the anomaly score exceeds the threshold, it is determined that the state corresponding to the sample is subject to GNSS interference. S5. Visualization of monitoring results: Visualize the GNSS interference monitoring results obtained in step S4.

[0008] Further, in step S1, the preprocessing includes filtering the validity of longitude, latitude, position navigation accuracy class (NACP), ground speed (GS), and track angle (TA) parameters; in step S2, the preset spatiotemporal grid is a 2°×2° latitude and longitude grid, divided at 10-minute intervals; the feature parameter set includes at least the statistical features of accuracy class (NACP), ground speed (GS), and track angle (TA).

[0009] Furthermore, in step S2, after extracting the feature parameter set, the method further includes constructing a comprehensive anomaly dimension index T, which is calculated using the following formula:

[0010] In steps S3 and S4, the isolated forest model is trained and detected using the comprehensive anomaly dimension index T as input features.

[0011] Furthermore, in step S4, the interference determination threshold τ is determined based on the distribution of abnormal scores in historical GNSS interference event data, specifically by selecting a specified high quantile value from the historical abnormal score distribution.

[0012] Furthermore, after step S4, step S6, multi-source data cross-verification, is included: using air traffic control QAM data or ground GNSS monitoring station data, the GNSS interference anomaly data determined in step S4 is verified.

[0013] Further, in step S3, constructing the isolated forest model includes setting a time window to divide the data into blocks and dynamically determining the size N of the isolated forest. The method for dynamically determining the size N is as follows:

[0014] in, This represents the amount of ADS-B data within the current time window.

[0015] Furthermore, the process of constructing the isolated forest model described in step S3 includes: S31. Initialize the root node and include all samples in the training subset into this node; S32. At the current node, randomly select a feature and randomly select a partitioning threshold θ within its value range; S33. Based on whether the feature value of the sample is less than θ, divide it into the left subtree or the right subtree of the current node; S34. Recursively repeat steps S32-S33 on each newly generated child node until any of the following termination conditions are met: the number of samples in the node is 1, all samples in the node have the same feature value, or the tree reaches a preset depth. S35. Repeat the above process to construct multiple independent isolated trees to form an isolated forest.

[0016] On the other hand, this application claims protection for a GNSS interference monitoring system based on ADS-B data and the isolated forest algorithm, for implementing the method described in any one of the foregoing statements, the system comprising: The data preprocessing module is used to collect raw ADS-B data and preprocess it to remove non-interference and abnormal data. The feature extraction and segmentation module is used to divide the preprocessed data according to a spatiotemporal grid and extract the feature parameter set; The Isolation Forest Modeling Module is used to build and train Isolation Forest models based on feature parameter sets. The interference detection module is used to calculate the anomaly score of the data samples using a trained isolated forest model, and to determine GNSS interference by comparing it with a threshold. The visualization module is used to present the GNSS interference monitoring results in a visual format.

[0017] Furthermore, it also includes a multi-source verification module, which is used to cross-verify interference detection results by combining air traffic control QAM data or ground GNSS monitoring station data.

[0018] Furthermore, the comprehensive visualization platform constructed by the visualization display module includes: The heat map generation unit is used to generate airspace interference heat maps that reflect the severity of interference in different areas based on geographic information. The trajectory drawing unit is used to draw flight trajectories on electronic maps and to mark normal and interfered segments in different styles; The time series comparison unit is used to display the normal and abnormal comparison curves of multiple parameters on a time series side by side.

[0019] Compared with the prior art, the present invention achieves the following beneficial technical effects: This invention achieves efficient monitoring of satellite navigation interference by integrating aviation broadcast data with intelligent anomaly detection algorithms. It can rapidly process massive amounts of flight data, accurately distinguish anomalies caused by interference from data deviations caused by other factors, and clearly display the interference situation through an intuitive visualization interface. This significantly improves the civil aviation sector's ability to detect, locate, and understand such safety hazards, thereby enhancing operational safety. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the GNSS interference monitoring method based on ADS-B and the isolated forest algorithm of the present invention.

[0022] Figure 2 This is a flowchart illustrating the process of establishing multiple isolated trees in an embodiment of the present invention.

[0023] Figure 3 This is a block diagram of the GNSS interference monitoring system based on ADS-B data and the isolated forest algorithm of this invention. Detailed Implementation

[0024] 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. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The GNSS interference monitoring method and system based on ADS-B and the isolated forest algorithm provided in this invention focuses on building a complete technical chain from data cleaning, feature engineering, intelligent detection, and intuitive display to achieve efficient, accurate, and visualized monitoring of civil aviation GNSS interference. The specific embodiments of this invention will be described in detail below with reference to the accompanying drawings.

[0026] The core process of the method of this invention is as follows: Figure 1 As shown, the main steps include: First, perform S1: ADS-B Data Acquisition and Preprocessing. Continuously collect raw ADS-B messages broadcast by aircraft within the target airspace using a network of ADS-B receivers deployed at ground stations. This raw data contains outliers caused by non-GNSS interference factors such as airborne equipment malfunctions, communication link packet loss, and message parsing errors. To improve the accuracy of subsequent monitoring, this "noise" data must first be cleaned. Preprocessing operations include, but are not limited to: verifying and removing data with latitude, longitude, and altitude parameters exceeding physically reasonable ranges; filtering out illegal values ​​for parameters such as NACP according to relevant protocols (e.g., DO-260B); and removing abnormal records such as ground speed and track angle that clearly do not conform to flight conditions, based on aircraft performance. This step yields a high-quality preprocessed dataset with obvious non-interference anomalies removed.

[0027] Secondly, S2 is executed: data spatiotemporal partitioning and feature extraction. Considering the vast airspace and continuous data, the preprocessed dataset is partitioned according to preset rules for ease of management and analysis. For example, a regular grid can be constructed in geographic space, and a fixed window can be defined on the time axis. Then, for each partitioned spatiotemporal unit (such as data from a grid within a specific time period), a set of core parameters reflecting GNSS signal quality and flight status is extracted as the basic features for subsequent analysis. These feature parameters are directly or indirectly derived from GNSS positioning results, and their abnormal fluctuations are closely related to GNSS interference.

[0028] Next, proceed to S3: Isolation Forest Model Construction and Training. Using the feature parameter set extracted in the previous step as training data, the Isolation Forest unsupervised learning algorithm, suitable for anomaly detection in high-dimensional data, is employed to construct the monitoring model.

[0029] like Figure 2 As shown, the core of the construction process is to build multiple isolated binary trees (Isolation Trees). The specific steps are as follows: S31 initializes the root node by including all samples from the training subset into the node: using the training subset of a single tree (a subset of samples randomly drawn without replacement from the total data) as input, initializes the root node and includes all samples into the node.

[0030] S32 At the current node, randomly select a feature and randomly select a partitioning threshold θ within its value range: In the data contained in the current node, randomly select a feature (in this invention, it is usually the constructed comprehensive anomaly dimension index T), and randomly extract a partitioning threshold θ between the minimum and maximum values ​​of the current value of the feature according to a uniform distribution.

[0031] S33 determines whether a sample's feature value is less than θ and assigns it to the left or right subtree of the current node: samples with feature values ​​less than θ are assigned to the left child node, and samples with feature values ​​greater than or equal to θ are assigned to the right child node.

[0032] S34 recursively repeats steps S32-S33 on each newly generated child node until one of the following termination conditions is met: the number of samples in the node is 1, all samples in the node have the same feature value, or the tree reaches a preset depth: Determine if the newly generated child node satisfies any of the following termination conditions: 1) The number of samples contained in this node is equal to 1; 2) All samples in this node have the same feature values; 3) The binary tree has reached the preset maximum depth (e.g., set to 8 levels). If this is satisfied, the node is marked as a leaf node, and further splitting stops.

[0033] S35 Repeat the above process to construct multiple independent isolated trees to form an isolated forest: For the left and right child nodes that do not meet the termination condition, repeat the above steps S31 to S34 respectively, and use them as the new current nodes to continue random feature selection and threshold division until all branches meet the termination condition, thereby completing the construction of an isolated tree.

[0034] By repeating the above process, a predetermined number of independent isolated trees are constructed, eventually forming an "isolated forest". Since outlier data points are more easily isolated by random partitioning in the algorithm's principle, their path length in the tree is usually shorter.

[0035] Next, S4: GNSS Interference Anomaly Detection is performed. The real-time received or historically stored ADS-B data to be monitored is input into the pre-trained Isolation Forest model. The model calculates an "anomaly score" for each data sample, which quantifies the likelihood that the sample is considered an anomaly. The higher the score, the greater the difference in characteristics between the sample and most normal flight data, and the more likely it is to be an anomaly. To identify specific anomalies caused by GNSS interference from among many anomalies, a reasonable interference judgment threshold τ needs to be set. When the anomaly score of a sample exceeds this interference judgment threshold τ, it is determined that the aircraft corresponding to that sample may have been subjected to GNSS interference at that moment.

[0036] Next, execute S4: Implement GNSS interference anomaly detection. Input the real-time received or historically stored ADS-B data to be monitored (converted to a comprehensive anomaly dimension index T) into the trained isolated forest model. For each input sample x, the model calculates the path length h from the root node to the isolated leaf node by traversing every isolated tree in the forest. The expected value of the path length of the sample across all trees is denoted as... Based on this, the anomaly score for this sample is calculated. :

[0037] in, This is a correction function for the average path length. This represents the number of training samples. The score... The value of τ ranges from 0 to 1. The closer the value is to 1, the higher the probability that the sample is considered an anomaly, i.e., the greater the difference in characteristics from most normal flight data. To distinguish specific anomalies caused by GNSS interference from numerous anomalies, a reasonable interference judgment threshold τ needs to be set. When the anomaly score of a sample... If the threshold is exceeded, it is determined that the aircraft corresponding to the sample may have been subjected to GNSS interference at that moment.

[0038] Finally, execute S5: Visualize the monitoring results. The GNSS interference anomalies automatically identified by the algorithm are displayed intuitively through graphical and chart-like methods. This can include presenting the airspace distribution of interference on an electronic map as a heatmap, drawing the trajectories of interfered flights using different colors and line types, or comparing changes in key parameters before and after interference using time-series graphs. Visualization transforms the abstract algorithm output into situational information easily understood by air traffic controllers, operations control personnel, and other non-algorithm experts, greatly enhancing the application value of the monitoring results.

[0039] Furthermore, in the preprocessing step S1, the validity screening is specifically implemented as follows: (a) Longitude: Only data with values ​​between -180° and 180° are retained; (b) Latitude: Only data with values ​​between -90° and 90° are retained; (c) Navigation Accuracy Class (NACP): Based on standards such as DO-260B, only valid values ​​between 0 and 11 are retained; (d) Ground Speed ​​(GS): Considering the operating characteristics of civil aviation transport aircraft, only data in the range of 0 to 600 knots are retained; (e) Track Angle (TA): Only valid angle values ​​in the range of 0° to 360° are retained. Through the hard screening of these five key parameters, a large number of invalid or obviously erroneous messages can be efficiently eliminated.

[0040] In the spatiotemporal partitioning of step S2, a preferred implementation of the preset spatiotemporal grid is as follows: spatially, the global or target airspace is divided into multiple grids of size 2° (latitude) × 2° (longitude); temporally, the continuous data stream is sliced ​​at a basic interval of 10 minutes. In this way, ADS-B data at any given time can be assigned to a specific "2° × 2° - 10-minute" spatiotemporal unit for analysis. The feature parameter set includes at least the statistical characteristics of NACP, GS, and TA. For example, the mean, standard deviation, and extreme values ​​of these parameters within each spatiotemporal unit can be calculated to describe the overall GNSS status and motion characteristics of the aircraft group within that unit.

[0041] Following feature extraction in step S2, to provide the isolated forest model with a more discriminative single-dimensional input, this invention constructs a comprehensive anomaly dimension index T. This index is obtained by weighted fusion of three key parameters: NACP, GS, and TA. Its calculation formula is as follows:

[0042] In this formula, NACP (Navigation Accuracy) is given the highest weight (0.8) because it is the most direct and sensitive representation of GNSS signal quality; a decrease in its value (deterioration in accuracy) is a strong indication of interference. Ground speed (GS) and track angle (TA) each have a weight of 0.1, used to capture kinematic anomalies caused by positioning errors. The design of "1 / (NACP+1)" in the formula means that the lower the NACP value (the worse the accuracy), the greater the contribution of this term, and thus the larger the T value, the more likely it is to indicate anomalies. Subsequent steps S3 (model training) and S4 (anomaly detection) both use T as the core input feature for calculation and judgment.

[0043] Regarding the interference detection threshold τ, its determination in practice relies on historical data. ADS-B data from confirmed historical GNSS interference events is collected, and the trained model is used to calculate the anomaly scores of these "interference samples" following the steps described above. This forms a set of historical interference anomaly score sets. The threshold τ can be a specified high quantile value of the score set, such as the 95th quantile.

[0044] This means that historically, 95% of known interference samples had anomaly scores below this value. Using τ as a threshold for new data can effectively detect most GNSS interference anomalies while controlling the false alarm rate.

[0045] Preferably, after an anomaly is identified in step S4, step S6, multi-source data cross-validation, can be implemented to further improve accuracy. Specifically, the GNSS interference anomaly event identified by the algorithm (including time, location, affected flights, etc.) is compared and verified with information from other independent data sources. These data sources may include: 1) air traffic control QAM data, which may contain anomaly events or communication interruptions recorded by air traffic controllers; 2) ground GNSS monitoring station data, which can directly monitor GNSS signal quality degradation in local areas. If the algorithm alarm highly matches these external data sources in time and space, the credibility of the alarm being caused by real GNSS interference is greatly enhanced, helping to eliminate misjudgments caused by severe aircraft maneuvers, occasional ADS-B equipment malfunctions, etc.

[0046] Regarding the method for dynamically determining the size N of the isolated forest involved in S3, the specific implementation during model construction in step S3 is as follows: First, a "time window segmentation" strategy is adopted to divide the data into 10-minute windows. The number of ADS-B messages contained in each data block within a time window is dynamically changing. Therefore, the required forest size N for constructing the isolated forest is dynamically determined by the following formula:

[0047] This represents the amount of ADS-B data within the current time window.

[0048] This formula ensures that the model complexity can adapt to the size of the data. When the data volume is small ( (Small), the forest size is close to 100 basic trees, ensuring basic detection capabilities; when the data volume is huge ( (The forest area is very large), and the forest size will increase linearly accordingly to maintain the ability to effectively model massive amounts of data and ensure detection stability. This dynamic mechanism optimizes the utilization efficiency of computing resources while ensuring monitoring effectiveness.

[0049] This application also discloses a GNSS interference monitoring system based on ADS-B data and the isolated forest algorithm, used to implement the aforementioned method, such as... Figure 3 As shown, the system includes: Data preprocessing module: Used to execute step S1, responsible for accessing the ADS-B data stream, cleaning the data according to preset rules (such as the aforementioned legality filtering rules), and outputting a high-quality dataset.

[0050] Feature extraction and segmentation module: used to execute step S2, responsible for organizing the preprocessed data according to the spatiotemporal grid (e.g., 2°×2°, 10 minutes), calculating the feature parameter set, and constructing the comprehensive anomaly dimension index T when necessary.

[0051] The isolated forest modeling module is used to execute step S3, and is responsible for configuring algorithm parameters (such as dynamically determining the forest size N), building, training and maintaining the isolated forest model.

[0052] Interference detection module: Used to execute step S4, responsible for inputting real-time data into the model, calculating the anomaly score, comparing it with the threshold τ, and outputting a GNSS interference alarm.

[0053] Visualization module: Used to execute step S5, responsible for calling the integrated visualization platform to present the detection results graphically in various forms such as heat maps, trajectory maps, and time series maps.

[0054] Multi-source verification module: As a preferred component, this module connects to external data sources such as air traffic control QAM and ground GNSS monitoring stations to cross-verify the output of the interference detection module, thereby improving the reliability of alarms.

[0055] The aforementioned visualization module is implemented on a comprehensive visualization platform. This platform integrates multiple views to provide intuitive insights for users with different needs: The heatmap generation unit (used to generate airspace interference heatmaps) uses a civil aviation electronic map as a base map, subdividing the airspace into 1km × 1km grids. It calculates the proportion of anomalous data or the average anomalous score for each grid within a selected time period and renders it using a gradient from green (normal), yellow (slight), orange (moderate) to red (severe). This view globally displays the severity and spatial clustering of interference.

[0056] The trajectory drawing unit (used to generate spatiotemporal trajectory maps) draws the complete flight trajectories of single or multiple flights on an electronic map. The trajectory lines are stylized based on the GNSS interference assessment results: normal segments are represented by thin blue lines, slightly interfered segments by thick yellow lines, and heavily interfered segments by thick red lines. Interactive markers are added at the interference points; clicking on them displays detailed parameter anomalies at that moment.

[0057] The time-series comparison unit (used to generate multi-parameter time-series comparison plots) displays the curves of multiple key parameters such as NACP, ground speed, track angle, latitude and longitude over time side by side within a cooperative view. It typically compares the normal reference curve with the actual curve during interference, and highlights the interference period with shading to clearly reveal the cooperative anomaly patterns of each parameter when interference occurs.

[0058] In summary, this invention, through rigorous data preprocessing, innovative feature engineering, efficient isolated forest algorithm, and rich visualization methods, forms a practical and effective GNSS interference monitoring solution.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A GNSS interference monitoring method based on ADS-B and the isolated forest algorithm, characterized in that, Includes the following steps: S1. ADS-B Data Acquisition and Preprocessing: Acquire raw ADS-B data in the target airspace; preprocess the raw ADS-B data to remove non-interference abnormal data caused by equipment failure or transmission errors, and obtain the preprocessed dataset; S2. Data Spatiotemporal Division and Feature Extraction: The preprocessed dataset is divided according to a preset spatiotemporal grid; for the data in each spatiotemporal grid, a set of feature parameters related to GNSS signal quality is extracted. S3. Construction and training of the isolated forest model: Based on the feature parameter set obtained in step S2, construct the isolated forest model; S4. GNSS Interference Anomaly Detection: Input the ADS-B data to be monitored into the trained isolated forest model and calculate the anomaly score for each data sample; compare the anomaly score with a preset interference judgment threshold τ. If the anomaly score exceeds the threshold, it is determined that the state corresponding to the sample is subject to GNSS interference. S5. Visualization of monitoring results: Visualize the GNSS interference monitoring results obtained in step S4.

2. The method according to claim 1, characterized in that, In step S1, the preprocessing includes filtering the validity of longitude, latitude, position navigation accuracy class (NACP), ground speed (GS), and track angle (TA) parameters; in step S2, the preset spatiotemporal grid is a 2°×2° latitude and longitude grid, divided at 10-minute intervals; the feature parameter set includes at least the statistical features of accuracy class (NACP), ground speed (GS), and track angle (TA).

3. The method according to claim 2, characterized in that, In step S2, after extracting the feature parameter set, the method further includes constructing a comprehensive anomaly dimension index T, which is calculated using the following formula: In steps S3 and S4, the isolated forest model is trained and detected using the comprehensive anomaly dimension index T as input features.

4. The method according to claim 1 or 3, characterized in that, In step S4, the interference determination threshold τ is determined based on the distribution of abnormal scores in historical GNSS interference event data, specifically by selecting a specified high quantile value from the historical abnormal score distribution.

5. The method according to claim 1, characterized in that, Step S4 is followed by step S6, multi-source data cross-verification: using air traffic control QAM data or ground GNSS monitoring station data to verify the GNSS interference anomaly data determined in step S4.

6. The method according to claim 1, characterized in that, In step S3, constructing the isolated forest model includes setting a time window to divide the data into blocks and dynamically determining the isolated forest size N. The method for dynamically determining the size N is as follows: in, This represents the amount of ADS-B data within the current time window.

7. The method according to claim 1, characterized in that, The process of constructing the isolated forest model in step S3 includes: S31. Initialize the root node and include all samples in the training subset into this node; S32. At the current node, randomly select a feature and randomly select a partitioning threshold θ within its value range; S33. Based on whether the feature value of the sample is less than θ, divide it into the left subtree or the right subtree of the current node; S34. Recursively repeat steps S32-S33 on each newly generated child node until any of the following termination conditions are met: the number of node samples is 1, all samples in the node have the same feature value, or the tree reaches a preset depth. S35. Repeat the above process to construct multiple independent isolated trees to form an isolated forest.

8. A GNSS interference monitoring system based on ADS-B data and the isolated forest algorithm, characterized in that, The system for implementing the method of any one of claims 1-7 comprises: The data preprocessing module is used to collect raw ADS-B data and preprocess it to remove non-interference and abnormal data. The feature extraction and segmentation module is used to divide the preprocessed data according to a spatiotemporal grid and extract the feature parameter set; The Isolation Forest Modeling Module is used to build and train Isolation Forest models based on feature parameter sets. The interference detection module is used to calculate the anomaly score of the data samples using a trained isolated forest model, and to determine GNSS interference by comparing it with a threshold. The visualization module is used to present the GNSS interference monitoring results in a visual format.

9. The system according to claim 8, characterized in that, It also includes a multi-source verification module, which is used to cross-verify interference detection results by combining air traffic control QAM data or ground GNSS monitoring station data.

10. The system according to claim 8 or 9, characterized in that, The comprehensive visualization platform constructed by the visualization module includes: The heat map generation unit is used to generate airspace interference heat maps that reflect the severity of interference in different areas based on geographic information. The trajectory drawing unit is used to draw flight trajectories on electronic maps and to mark normal and interfered segments in different styles; The time series comparison unit is used to display the normal and abnormal comparison curves of multiple parameters on a time series side by side.

Images