Sensitive data exploration method and system based on machine learning

Abstract

The invention relates to a sensitive data exploration method and system based on machine learning, and the method comprises the steps: collecting original data including tower position data, airport position data and important station position data from a power related data source, carrying out the preprocessing operation of abnormal value cleaning, data format unification and the like, and generating structured data; based on the preprocessed data, multi-dimensional feature vectors including position distribution, spatial relevance and data frequency are extracted, model training data are generated in combination with a sensitivity labeling rule, and a machine learning algorithm is utilized to construct and optimize a sensitive data exploration model; a target data set is subjected to sensitivity analysis through a trained model, sensitive classification labels are generated, probed sensitive data are subjected to hierarchical protection according to the classification labels, and encryption storage and access control strategies are adopted for high-sensitivity data; the sensitive data can be efficiently and accurately identified and protected, and the method is suitable for data security requirements in the field related to electric power.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and is a sensitive data detection method and system based on machine learning. Background Art

[0002] With the rapid development of information technology, the ability to collect, store and analyze data has been continuously improved. In the power-related field, data-driven decision-making and management have become an important means to improve efficiency and safety. Traditionally, power data mainly exists in the form of regional distribution, including geographic information such as tower locations, airport locations and important station locations. These data are of great value in power network design, operation optimization and risk assessment, but also face severe challenges in data protection and sensitivity identification.

[0003] However, existing technologies generally lack the ability to efficiently identify and protect sensitive data when processing this data, and usually rely on manual rules or static classification methods, which are inefficient when the data dimensions are complex and the characteristics are diverse. In addition, these traditional methods cannot dynamically adapt to changes in data distribution, and it is difficult to formulate accurate protection strategies for different sensitivity levels, resulting in the risk of leakage or insufficient protection of sensitive data during processing.

[0004] Therefore, it is necessary to develop a new sensitive data detection method and system based on machine learning. Summary of the invention

[0005] The present application provides a sensitive data detection method and system based on machine learning to improve the efficiency and accuracy of sensitive data identification.

[0006] The present application provides a sensitive data detection method based on machine learning, comprising: Collecting raw data including tower location data, airport location data, and important station location data from power-related data sources; preprocessing the raw data, wherein the preprocessing includes cleaning abnormal values, unifying data formats, and standardizing geographic coordinates to generate preprocessed structured data; Based on the preprocessed structured data, a multidimensional feature vector is generated; according to a predetermined sensitivity labeling rule, the multidimensional feature vector is combined with its corresponding sensitivity labeling label to generate model training data; wherein the multidimensional feature vector includes location distribution, spatial correlation, and data frequency; Based on the model training data, a sensitive data detection model is constructed using a machine learning algorithm; the sensitive data detection model is trained by iterative optimization to obtain model parameters for identifying the sensitivity of target data; and a trained sensitive data detection model is obtained according to the model parameters; Input the target data set into the sensitive data detection model to obtain the model output result; according to the model output result, label the sensitive data included in the target data set to generate a sensitivity classification label; wherein the target data set uses the preprocessed target structured data; Based on the sensitivity classification labels, the detected sensitive data is protected in different levels, including encrypted storage and access control strategies for highly sensitive data.

[0007] Furthermore, the generating of a multi-dimensional feature vector based on the pre-processed structured data includes: Divide the input geographic data into equidistant grid cells according to a preset rule, wherein each grid cell corresponds to a specified geographic range; calculate the number of geographic data points in each grid cell, expressed as grid density; normalize the grid density value to generate a location distribution feature reflecting the local distribution feature of the geographic data; For each geographic data point, calculate its distance from a set of predefined important facilities; assign weights according to facility types, perform weighted average of each distance value and the corresponding weight to obtain a spatial correlation value; normalize the spatial correlation values ​​of all geographic data points to generate spatial correlation features; wherein the important facilities include towers, airports and substations; The time attributes of each geographic data point are analyzed, and the number of occurrences within a fixed time period is counted as the frequency value; by setting the time decay weight, the frequency value is weighted and normalized to generate the spatiotemporal frequency characteristics that describe the spatiotemporal activity level; The normalized location distribution features, spatial correlation features, and spatiotemporal frequency features of each geographic data point are combined to generate a multidimensional feature vector, and a feature vector dataset is output for model training.

[0008] Furthermore, for each geographic data point, the distance between it and a set of predefined important facilities is calculated; weights are assigned according to facility types, and each distance value and the corresponding weight are weighted averaged to obtain a spatial correlation value; the spatial correlation values ​​of all geographic data points are normalized to generate spatial correlation features, including: For each geographic data point , calculate its important facilities set Each facility distance, generate a distance matrix , among which, Line Column value Indicates geographic data points to The distance to the facility, Calculate according to the following formula (1): ; in, and Respectively geographic data points and The location coordinates of each facility; According to Risk factor of important facilities , a dynamic weight is assigned to each facility according to the following formula (2): ; in, For the The weight coefficient of each important facility; The number of important facilities; According to the following formula (3), the spatial association value of the geographic data point is calculated : ; in, is the attenuation coefficient, which is used to control the effect of distance on spatial correlation; For the matrix No. The maximum distance value of the column, used for normalized distance; The number of important facilities; According to the following formula (4), the spatial correlation value of all geographic data points is Perform normalization: ; in, Normalized spatial correlation value; For all spatial association values The minimum value in ; For all spatial association values The maximum value in ; A spatial correlation feature is generated according to the normalized spatial correlation value.

[0009] The beneficial effects of the technical solution provided by this application include: (1) The present invention uses a machine learning-based sensitive data detection model, multi-dimensional feature vectors and sensitivity labeling rules to achieve automatic classification and accurate identification of target data, significantly improving the efficiency and accuracy of sensitive data identification. (2) By iteratively optimizing the training model parameters, the present invention can dynamically adjust the detection model according to changes in data distribution, ensuring the applicability of the model in a variety of scenarios and data complexity conditions, thereby avoiding the limitations of traditional static classification methods. (3) The present invention introduces multi-dimensional features such as location distribution, spatial correlation and data frequency, and comprehensively models the spatial and temporal correlation characteristics of model data, providing a more complete expression of data sensitivity and solving the problem of insufficient single-dimensional analysis. (4) The present invention performs hierarchical protection of data according to sensitivity classification labels, especially using encrypted storage and strict access control strategies for highly sensitive data, thereby achieving targeted protection of sensitive data and enhancing data security. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a flowchart of a sensitive data detection method based on machine learning provided in the first embodiment of the present application.

[0011] Figure 2 It is a schematic diagram of a sensitive data detection system based on machine learning provided in the second embodiment of the present application. DETAILED DESCRIPTION

[0012] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0013] The first embodiment of the present application provides a sensitive data detection method based on machine learning. Figure 1 , which is a schematic diagram of the first embodiment of the present application. Figure 1 A sensitive data detection method based on machine learning is described in detail in the first embodiment of the present application.

[0014] Step S101: Collecting raw data including tower location data, airport location data, and important station location data from power-related data sources; preprocessing the raw data, wherein the preprocessing includes cleaning abnormal values, unifying data formats, and standardizing geographic coordinates to generate preprocessed structured data.

[0015] Step S101 includes two main processes: raw data collection and preprocessing. These two processes are described in detail below.

[0016] First, collect raw data including tower location data, airport location data and important station location data from power-related data sources. The data source can be a distributed sensor network, a geographic information system (GIS), a satellite remote sensing system or a manually maintained database. The collection process can be carried out through programming interface calls, real-time data streams or batch imports. Taking tower location data as an example, the longitude and latitude coordinate information of the tower can be exported through the GIS platform, along with other geographic attributes, such as the tower type and the area number where it is located. Airport location data can be obtained through the aviation management database, including the precise location of the airport, classification information (such as international airports, local airports) and distribution information of nearby facilities. Important station location data can be extracted from the internal database of the power company. The data format may be CSV, JSON or XML files, including the station name, geographic coordinates and its functional attributes.

[0017] After the collection is completed, the preprocessing stage begins. First, the raw data is cleaned of outliers. Statistical methods can be used to identify outliers, such as performing a range check on the location coordinates to remove points that exceed reasonable geographic boundaries; and removing or correcting entries with incorrect data types (such as records whose latitude and longitude formats do not meet the standards). During the cleaning process, it is necessary to ensure that the integrity and accuracy of the data are retained. For example, when some records are missing, they can be reasonably filled in through interpolation or statistical distribution of data from adjacent areas.

[0018] Secondly, unifying the data format is a key step. In this process, data from different data sources are standardized into a unified file format, such as converting JSON or XML format into a tabular CSV file, or directly storing it as a table structure of a relational database. In the process of unifying the data format, ensure that each record has a consistent field structure, for example, each record contains fields such as "data type", "geographic coordinates", and "timestamp".

[0019] Then, the geographic coordinates are normalized.

[0020] Finally, the preprocessed data is stored as structured data. You can use a relational database (such as MySQL, PostgreSQL) to store each record, or a distributed storage system (such as Hadoop HDFS) to manage large data sets. Each record is represented in the database with standardized fields, such as "data type" is "tower", "airport" or "station", "geographic coordinates" are represented by latitude and longitude values, and "additional attributes" store specific functional information, such as the load capacity of the tower or the service range of the station.

[0021] Step S102: Generate a multidimensional feature vector based on the preprocessed structured data; combine the multidimensional feature vector with its corresponding sensitivity label according to a predetermined sensitivity labeling rule to generate model training data; wherein the multidimensional feature vector includes position distribution, spatial correlation, and data frequency.

[0022] In step S102, key information is first extracted from the preprocessed structured data generated in step S101, and a multidimensional feature vector is constructed based on the key information to describe the data features. The generation process of the multidimensional feature vector includes three main aspects: location distribution features, spatial correlation features, and data frequency features.

[0023] The extraction of location distribution features is completed by analyzing the distribution of data points in geographic space. The geographical location of the data points is divided into several regions according to preset rules. The division of regions can be a fixed grid or a dynamic division based on geographical characteristics. The distribution density of data points is calculated in each region, and the statistical results are standardized to reflect the spatial distribution characteristics of data points in different regions. This process ensures that the generated features can accurately describe the local distribution pattern of data points while maintaining consistency with the distribution in the entire spatial range.

[0024] The extraction of spatial correlation features requires the analysis of the spatial relationship between data points and between data points and specific facilities (such as towers, airports, and important stations). For each data point, the distance to the target facility is calculated, and a weight is assigned based on the type and importance of the target facility to reflect the strength of the spatial correlation. Then, these distance and weight information are combined into a set of feature values ​​that can quantify spatial correlation. These feature values ​​can describe the relative spatial position relationship between data points and facilities, thereby capturing potential geographic correlation.

[0025] The extraction of data frequency features is achieved by analyzing the time attributes of data points. The time attributes of data points are divided into several fixed periods, such as counting the number of occurrences of each data point in the corresponding period by day, week or month, so as to construct basic frequency features. On this basis, the frequency value of the data point is adjusted by introducing the time decay weight, so that the activity in the recent time has a greater impact on the feature value, which is more in line with the impact of time changes on data sensitivity in actual scenarios.

[0026] After generating location distribution features, spatial correlation features, and data frequency features, these features are combined into a multidimensional feature vector, where each dimension represents an independent feature value. The generated multidimensional feature vector fully reflects the spatial distribution, temporal activity, and correlation between data points and facilities, providing a basis for subsequent sensitivity analysis.

[0027] Next, the generated multidimensional feature vectors are combined with their corresponding sensitivity annotation labels to construct model training data according to pre-defined annotation rules. Annotation rules can be formulated based on the geographic attributes of data points, their relationship with target facilities, and historical experience knowledge. By combining multidimensional feature vectors with sensitivity labels, the generated model training data has rich feature information and clear target annotations, laying a solid foundation for subsequent model construction and optimization.

[0028] Furthermore, the generating of a multi-dimensional feature vector based on the pre-processed structured data includes: Divide the input geographic data into equidistant grid cells according to a preset rule, wherein each grid cell corresponds to a specified geographic range; calculate the number of geographic data points in each grid cell, expressed as grid density; normalize the grid density value to generate a location distribution feature reflecting the local distribution feature of the geographic data; For each geographic data point, calculate its distance from a set of predefined important facilities; assign weights according to facility types, perform weighted average of each distance value and the corresponding weight to obtain a spatial correlation value; normalize the spatial correlation values ​​of all geographic data points to generate spatial correlation features; wherein the important facilities include towers, airports and substations; The time attributes of each geographic data point are analyzed, and the number of occurrences within a fixed time period is counted as the frequency value; by setting the time decay weight, the frequency value is weighted and normalized to generate the spatiotemporal frequency characteristics that describe the spatiotemporal activity level; The normalized location distribution features, spatial correlation features, and spatiotemporal frequency features of each geographic data point are combined to generate a multidimensional feature vector, and a feature vector dataset is output for model training.

[0029] In this embodiment, the process of generating a multi-dimensional feature vector through preprocessed structured data can be divided into several detailed steps, each of which has a clear goal and method to ensure that the core features of the data can be fully and accurately extracted to provide high-quality input for the subsequent training of the machine learning model.

[0030] First, based on the input geographic data, the target area is divided according to preset rules. This division is completed based on equidistant grid cells, and each grid cell covers a specified geographic range. The purpose of grid division is to classify geographic data points into individual cells in order to analyze the distribution characteristics of data points in the local area. After the division is completed, the geographic data points in each grid cell are counted, and the number of data points contained in each cell is recorded and expressed as a grid density value. In order to eliminate the influence of the absolute number difference between the grid density values ​​on the subsequent analysis, all grid density values ​​are normalized to ensure that these values ​​are distributed within a fixed range, such as standardizing them to between 0 and 1. This normalization process enables the location distribution characteristics to more intuitively reflect the density of geographic data points in different regions, and generate location distribution characteristics of local distribution characteristics.

[0031] Next, in order to further analyze the spatial correlation between geographic data points, it is necessary to calculate the distance between each data point and a set of predefined important facilities. Important facilities can include poles, airports, and substations, which are defined as target points with special significance, reflecting the potential correlation between geographic data points and critical infrastructure. For each geographic data point, its straight-line distance to each important facility is calculated, and a specific weight is assigned according to the type of facility. For example, poles may have a lower weight, while substations have a higher weight. After the calculation is completed, each distance value is weighted averaged with the corresponding weight to generate a spatial correlation value for each geographic data point. In order to ensure the comparability between these correlation values, the spatial correlation values ​​of all data points are normalized to generate a spatial correlation feature. This feature can fully describe the relative relationship and potential connection between geographic data points and target facilities.

[0032] Subsequently, in order to capture the activity of geographic data points in the time dimension, it is necessary to analyze the time attributes of each data point. First, the time attributes are divided into fixed time periods, such as days, weeks or months, and the number of occurrences of data points in each time period is counted and expressed as frequency values. These frequency values ​​reflect the activity of data points in different time periods. On this basis, the frequency values ​​are weighted by introducing time decay weights, so that the activity in the recent time period contributes more to the frequency characteristics, while the influence of the long-term time period gradually decreases. This processing method can more realistically reflect the spatiotemporal dynamic characteristics of data points. Finally, the weighted frequency values ​​are normalized to generate spatiotemporal frequency characteristics that describe the spatiotemporal activity of data points.

[0033] After the above feature extraction steps are completed, the normalized location distribution features, spatial correlation features, and spatiotemporal frequency features of each geographic data point are combined to generate a multidimensional feature vector. The dimensions of each feature vector correspond to the specific components of these features, thereby fully reflecting the local distribution, spatial correlation, and spatiotemporal activity of the geographic data points. The generated multidimensional feature vector is then output as a data set for model training, providing high-quality input data for the machine learning algorithm in the subsequent steps.

[0034] Furthermore, for each geographic data point, the distance between it and a set of predefined important facilities is calculated; weights are assigned according to facility types, and each distance value and the corresponding weight are weighted averaged to obtain a spatial correlation value; the spatial correlation values ​​of all geographic data points are normalized to generate spatial correlation features, including: For each geographic data point , calculate its important facilities set Each facility distance, generate a distance matrix .in, is the number of important facilities. The set of important facilities includes facilities such as towers, airports and substations. The geographical location of each facility is expressed in coordinate form. The coordinates of each geographic data point P are , each facility The coordinates of .

[0035] Among them, Line Column value Indicates geographic data points to The distance to the facility, Calculate according to the following formula (1): ; in, and Respectively geographic data points and The location coordinates of each facility; Next, based on the risk factor of the facility Assign a dynamic weight to each facility. Risk Factor is a predefined parameter used to measure the importance of a facility. For example, a tower may have a lower risk factor (such as 0.5), while an airport or substation may have a higher risk factor (such as 0.8 or 1.0). The specific value can be set based on historical data analysis or business needs. Formula (2) defines how dynamic weights are allocated: ; in, For the The weight coefficient of each important facility; The number of important facilities; According to the following formula (3), the spatial association value of the geographic data point is calculated : ; in, is the attenuation coefficient, which is used to control the effect of distance on spatial correlation; For the matrix No. The maximum distance value of the column, used for normalized distance; The number of important facilities; In formula 2, is the distance attenuation coefficient, which is used to control the effect of distance on spatial correlation values. Its value range is usually between 0.1 and 1. The value indicates that the influence of distance is more significant, for example, when When the facilities are close to The contribution increased significantly. is a matrix No. The maximum distance value of the column, used for distance Standardize the distances between facilities so that they are comparable. The maximum distance between columns is 50, and , then the standardized distance is The weighted summation in the formula ensures that the contribution of each facility is adjusted by weight and distance decay, generating the initial spatial association value of the geographic data points.

[0036] According to the following formula (4), the spatial correlation value of all geographic data points is Perform normalization: ; in, Normalized spatial correlation value; For all spatial association values The minimum value in ; For all spatial association values The maximum value in ; A spatial correlation feature is generated according to the normalized spatial correlation value.

[0037] Furthermore, the time attribute of each geographic data point is analyzed, and the number of occurrences within a fixed time period is counted as a frequency value; by setting a time decay weight, the frequency value is weighted and normalized to generate a spatiotemporal frequency feature describing the spatiotemporal activity, including: For geographic data points at fixed time periods The number of occurrences within , according to the decay weight at the current time point The time period can be defined according to the application scenario, for example, the frequency of occurrence of each data point can be counted in units of days, weeks or months. If statistics are selected by day, Indicates the number of times a data point is recorded in a certain day. For example, the number of times a data point P appears in the past 7 days may be {5, 3, 6, 2, 8, 4, 1}, corresponding to The values ​​of are these daily occurrences.

[0038] Next, according to the current time point and time period Time interval , calculate the attenuation weight by formula (5) The definition of decay weight is to control the impact of long-term data on frequency calculation, ensuring that recent data contributes more, while the impact of long-term data gradually weakens. Formula (5) is defined as: ; in, Indicates the current time and time period time interval; is the attenuation control parameter, which is used to adjust the weight impact of long-term data; It is an adjustable exponential parameter used to control the nonlinearity of attenuation.

[0039] Indicates the current time point and time period The time interval is usually calculated in days or hours. For example, the current time is November 28, 2024. For the records on November 21, 2024, the time interval 7 days. Parameters is the attenuation control parameter, which controls the impact of future data on the weight. The recommended value range is 0.1 to 1.0. The value makes the weight of the distant data decrease faster. is an adjustable exponential parameter used to control the nonlinearity of the attenuation, for example When the weight decays linearly, In this example, we can set and , when When , we can calculate .

[0040] In each time period According to the following formula (6), the weighted frequency value is calculated: : ; in, is the total number of time periods; is the weighted cumulative frequency value of the geographic data point in all time periods; according to the following formula (7), the normalized spatiotemporal frequency feature is generated : ; in, is the weighted frequency mean of all time periods; is the standard deviation of the weighted frequency values, which is used to measure the difference in activity of a data point relative to the overall distribution.

[0041] Step S103: Based on the model training data, a sensitive data detection model is constructed using a machine learning algorithm; the sensitive data detection model is trained through iterative optimization to obtain model parameters for identifying the sensitivity of the target data; and a trained sensitive data detection model is obtained based on the model parameters.

[0042] In step S103, based on the model training data generated in step S102, a sensitive data detection model is constructed using a machine learning algorithm, and the model is trained through iterative optimization to generate model parameters for identifying the sensitivity of the target data.

[0043] First, extract multidimensional feature vectors and their corresponding sensitivity annotation labels from the model training data. The feature vectors are used as the input of the model, and the labels are used as supervisory signals to guide the model's learning process. In the early stages of training, select an appropriate machine learning algorithm to build an initial model, such as a deep neural network, support vector machine, or random forest model, and set the initial model parameters. Taking a deep neural network as an example, the initial model can be composed of several layers of fully connected networks, and the number of nodes in each layer is designed according to the dimension of the feature vector and the complexity requirements of the model.

[0044] Next, the training data is input into the model in batches, the error between the sensitivity label predicted by the model and the true label is calculated, and the model parameters are adjusted according to the error. Error calculation can use common loss functions, such as the cross entropy loss function, to quantify the difference between the predicted value and the true label. Based on error feedback, the model parameters are updated using optimization algorithms (such as stochastic gradient descent or Adam optimization algorithm) so that the model's prediction results gradually approach the true label.

[0045] In order to further improve the generalization ability of the model, regularization techniques such as L2 regularization and Dropout method are introduced in the training process to reduce the risk of overfitting. In addition, the performance changes of the model are monitored by using the validation set to determine whether the training process tends to converge. When the loss on the validation set no longer decreases significantly, the training process is stopped and the current model parameters are saved.

[0046] After completing the model training, the model performance is evaluated, and the key indicators such as sensitivity recognition accuracy, recall rate and F1 score of the model are calculated using the test set to verify the model's sensitivity classification ability for unseen data. If the performance indicators do not meet expectations, the model can be retrained by adjusting the model structure, optimizing the algorithm or hyperparameters, etc. until the model meets the requirements.

[0047] Finally, the trained sensitive data detection model and its parameters are saved. The model can accurately predict the sensitivity label based on the feature vector of the input data and provide support for sensitive data detection in subsequent steps. The entire model construction and training process aims to obtain the optimal parameters to ensure that the model has stable prediction performance and efficient recognition capabilities.

[0048] Furthermore, the sensitive data detection model includes a feature extraction layer, a feature fusion layer, and a sensitivity assessment layer; The feature extraction layer is used to receive the pre-processed target structured data; the feature extraction layer uses a parallel spatial feature extraction branch and an attribute feature extraction branch to process the received data respectively; the spatial feature extraction branch is implemented using a spatial convolutional network, which is used to extract the position distribution characteristics and spatial correlation characteristics of each data point through multi-layer convolution to generate a spatial feature vector; the attribute feature extraction branch is implemented using an attention mechanism to capture the temporal correlation characteristics and activity level of the data point to generate an attribute feature vector; The feature fusion layer is used to receive the spatial feature vector and the attribute feature vector provided by the feature extraction layer; the feature fusion layer is implemented using a bidirectional gated recurrent unit network to process the received data and generate a unified comprehensive feature vector; The sensitivity assessment layer is used to receive the comprehensive feature vector provided by the feature fusion layer; the sensitivity assessment layer is used to input the comprehensive feature vector into a multi-layer fully connected network to calculate the sensitivity score of each data point; the scoring result is calibrated according to predefined sensitivity rules; based on the sensitivity score and rule calibration results, a dynamic confidence threshold is used to determine whether the data point belongs to a specific sensitivity category, and a sensitivity classification label is generated.

[0049] In this embodiment, the sensitive data detection model realizes efficient processing and sensitivity analysis of target data through the synergy of feature extraction layer, feature fusion layer and sensitivity assessment layer. Each component of the model has a clear function and implementation method, and the data flow and processing logic between each part are closely connected to ensure the efficient realization of the overall function.

[0050] First, the feature extraction layer receives the preprocessed target structured data, including the geographic location, time attributes, and other relevant features of the data points. The feature extraction layer processes the data through parallel spatial feature extraction branches and attribute feature extraction branches. The spatial feature extraction branch is implemented using a spatial convolutional network, which extracts the location distribution features and spatial correlation features of each data point through multi-layer convolution operations. Specifically, the spatial convolutional network takes the geographic location of the data point as input, captures the local distribution pattern of the data point through the convolution kernel of the local receptive field, and extracts the global correlation information between the data point and other regions through the layer-by-layer expanded receptive field. The output spatial feature vector fully reflects the distribution characteristics and correlation of the data points in the spatial dimension.

[0051] The attribute feature extraction branch is implemented using the attention mechanism to capture the temporal correlation characteristics and activity level of data points. The attention mechanism evaluates the key time points in the time series by dynamically allocating weights, thereby highlighting the importance of recent data and generating an attribute feature vector in combination with the time series context information. This feature vector can accurately describe the dynamic performance of data points in the time dimension, providing high-quality input for subsequent feature fusion.

[0052] The feature fusion layer is used to receive spatial feature vectors and attribute feature vectors, and realizes the fusion processing of these two types of features through a bidirectional gated recurrent unit network (GRU). The GRU network has high efficiency and robustness in processing time series data and multi-dimensional input features. Through the bidirectional structure, the network can simultaneously consider the forward and backward information of the time series, thereby capturing the complex associations between features. The fused output is a comprehensive feature vector that uniformly includes the spatial distribution characteristics of the data points, the time series dynamic characteristics, and other related characteristics.

[0053] After receiving the comprehensive feature vector, the sensitivity assessment layer uses a multi-layer fully connected network to calculate the sensitivity score of each data point. The fully connected network maps the comprehensive feature vector to a score value through layer-by-layer weight mapping and activation function, which represents the possibility that the data point belongs to a specific sensitivity category. Subsequently, the scoring results are calibrated according to predefined sensitivity rules, such as setting priorities or minimum scoring thresholds for specific categories. Finally, the sensitivity classification label of each data point is determined by judging the dynamic confidence threshold. The dynamic confidence threshold is adjusted according to the data distribution and application scenario to ensure the accuracy and applicability of the classification results.

[0054] Through the above process, this model realizes the complete data processing flow from feature extraction to classification labeling. It can efficiently identify sensitive data and label its sensitivity categories in complex and multi-dimensional data scenarios, providing strong support for subsequent data management and protection measures.

[0055] Furthermore, the spatial convolutional network used by the spatial feature extraction branch includes an input layer, a local receptive field convolutional layer, a cross-region convolutional layer, a multi-scale feature fusion layer, a normalization and output layer; The input layer is used to receive the preprocessed target structured data, organize the data containing the geographical location information and related regional features of each data point into a two-dimensional matrix form; each row of the two-dimensional matrix represents the spatial information of a data point, and each column represents the specific features of the data point, thereby generating an input feature matrix; The local receptive field convolution layer is used to process the input feature matrix, and by setting a convolution kernel with a fixed range, performs a convolution operation on the features of each data point and its neighboring data points, extracts the local spatial distribution characteristics, and generates a preliminary local feature map, which is used to describe the neighborhood distribution pattern of the data points and the feature distribution density of the local area; The cross-region convolution layer is used to receive the local feature map and extract the potential spatial correlation between data points in a large range by expanding the convolution operation of the receptive field range; using multiple convolution kernels of different sizes, cross-region features are extracted at different receptive field scales to generate a global spatial feature map to reflect the correlation and spatial characteristics between distant data points; The multi-scale feature fusion layer is used to receive the local feature map and the global spatial feature map, and fuse the two according to a specified weight ratio; the fused comprehensive feature map provides a unified description of the local and global relationship of the data points; The normalization and output layer is used to perform batch normalization on the comprehensive feature map, balance the numerical differences between the output features of different convolutional layers, and generate a normalized spatial feature vector.

[0056] In this embodiment, the spatial convolutional network of the spatial feature extraction branch realizes efficient processing and comprehensive feature extraction of geographic data with its hierarchical structure. The entire network consists of an input layer, a local receptive field convolution layer, a cross-region convolution layer, a multi-scale feature fusion layer, and a normalization and output layer. Each part is interconnected to ensure that the spatial information of the data point can be fully modeled.

[0057] First, the input layer is responsible for receiving the preprocessed target structured data and organizing it into a two-dimensional matrix. Each row corresponds to a geographic data point, representing the spatial information of the data point, such as geographic coordinates, area type, or related additional feature information. Each column represents a specific feature dimension, such as the relevance index of a specific facility or other numerical attributes of the data point. The input layer provides the basis for subsequent convolution operations by converting the raw data into an input feature matrix, ensuring that the data is transmitted in a structured form.

[0058] The local receptive field convolution layer processes the input feature matrix to extract local spatial distribution characteristics by setting a fixed range of convolution kernels. In this process, the size of the convolution kernel determines the range of the neighborhood covered in each operation. For example, a 3×3 convolution kernel can capture the spatial distribution pattern of a data point and its closest neighboring data points. Through the convolution operation, the local features of each area are calculated to generate a preliminary local feature map. The local feature map reflects the feature distribution density of data points in a small range, such as the point density in the neighborhood or the correlation between adjacent data points. The local receptive field convolution layer can capture fine spatial information and is suitable for analyzing feature differences in local areas.

[0059] The cross-region convolution layer further expands the receptive field to extract the potential global correlation between data points. This layer performs operations based on multiple convolution kernels of different sizes, such as 5×5 or 7×7 convolution kernels, to capture medium-range and larger-range spatial relationships, respectively. Through these convolution kernels of different scales, the cross-region convolution layer is able to generate a global spatial feature map that reflects the correlation between distant data points. For example, a larger convolution kernel can identify the relationship between two distant but potentially functionally related data points. The cross-region convolution layer effectively integrates the global spatial characteristics of the data and provides rich global information for subsequent feature fusion.

[0060] The multi-scale feature fusion layer is responsible for fusing the local feature map and the global spatial feature map according to the specified weight ratio. During the fusion process, the contribution ratio of local features and global features is dynamically adjusted through the weighted average method to meet the needs of specific application scenarios. For example, in some applications that emphasize details, local features may account for a larger proportion, while in scenarios where the overall trend needs to be considered, the weight of global features may be higher. The fused comprehensive feature map retains the neighborhood distribution information and global spatial pattern of the data points at the same time, providing a unified and comprehensive feature description.

[0061] Finally, the normalization and output layer performs batch normalization on the comprehensive feature map to balance the numerical differences between the output features of different convolutional layers. The normalization operation can adjust the distribution of feature values, such as scaling them to a specific range (such as 0 to 1 or standard normal distribution) to eliminate the impact of scale differences between feature dimensions on subsequent analysis. After normalization, the generated output is a normalized spatial feature vector, with each data point corresponding to a feature vector, which is used to describe its comprehensive characteristics in the spatial dimension.

[0062] Through the collaborative work of the above layers, the spatial convolutional network can efficiently process complex geographic data and comprehensively model the spatial information of data points from both local and global levels, providing high-quality input for subsequent machine learning models.

[0063] Furthermore, the attention mechanism of the attribute feature extraction branch includes the following implementation process: receiving the preprocessed target structured data, and representing the time attribute as a time series feature vector; Calculate the context weight for the feature vector at each time point and dynamically assign weights based on time differences and feature similarities; Use context weights to perform weighted aggregation on time series feature vectors to generate enhanced time feature vectors; All enhanced temporal feature vectors are integrated into a global temporal feature representation and output to the feature fusion layer for further processing.

[0064] In this embodiment, the attention mechanism of the attribute feature extraction branch is implemented through a series of clear steps, aiming to extract temporal features from the temporal attributes of the target data and generate high-quality temporal feature representations to provide support for subsequent feature fusion and sensitivity analysis. The entire process is based on the preprocessed target structured data, and the extraction and enhancement of temporal features are achieved through dynamic weight allocation and feature aggregation.

[0065] First, the preprocessed target structured data is received, and the time attributes of the data are extracted and represented as a time series feature vector. The time attribute usually includes the timestamp information of the data point, such as the time when an event occurred or the time when a geographic data point was recorded. By parsing the timestamp and combining other features of the data point, the relevant information of each time point is encoded as a feature vector. For example, if the characteristics of a data point in different time periods include the degree of activity, the number of associated facilities, and regional changes, this information can be represented as a multidimensional vector to form a time series feature set.

[0066] Next, the contextual weights are calculated for the feature vectors at each time point, a process that is the core of the attention mechanism. The calculation of contextual weights takes into account both the temporal difference and feature similarity between time points. The temporal difference represents the distance between different time points, measured in hours or days, for example, while feature similarity is quantified by comparing the values ​​of the feature vectors at the time points. Through this combination, the attention mechanism is able to dynamically assign weights to highlight contextual information that is more relevant to the current time point, while weakening distant or less relevant features. For example, when two time points are adjacent and have similar features, their contextual weights are higher, while time points that are farther apart or have larger feature differences have lower weights.

[0067] After the calculation of context weights is completed, these weights are used to perform weighted aggregation on the time series feature vectors. The purpose of weighted aggregation is to generate enhanced time feature vectors so that the features of each time point not only contain its own information, but also combine the key information in its context. For example, through the aggregation operation, the activity level of a time point depends not only on the feature value of the point, but also on the influence of other time points associated with it. This operation makes the generated feature vector more reflective of the global dynamics of the time series.

[0068] Finally, the enhanced feature vectors of all time points are integrated into a global temporal feature representation. This integration process is achieved by arranging or concatenating all feature vectors in the time series to form a unified temporal feature matrix or vector set. The global temporal feature representation not only retains the local temporal information of the data points, but also captures the dynamic patterns and long-term trends in the entire time series. The generated global temporal features are passed as output to the feature fusion layer, which is combined with the spatial features for further processing.

[0069] Through the above implementation process, the attention mechanism ensures that the extraction of time series features is highly adaptable and flexible. Its dynamic weight allocation mechanism can highlight the feature contribution of key time points based on the time and feature relationship of actual data, thereby providing accurate time series information support for sensitive data exploration models.

[0070] Furthermore, the bidirectional gated recurrent unit network used in the feature fusion layer jointly models the correlation and spatial distribution pattern of spatial feature vectors and attribute feature vectors in the time dimension by introducing a dynamic context-aware mechanism, wherein the dynamic context-aware mechanism dynamically adjusts the weights of temporal features according to the regional importance of spatial features.

[0071] In this embodiment, the feature fusion layer uses a bidirectional gated recurrent unit network (GRU) combined with a dynamic context-aware mechanism to jointly model the input spatial feature vector and attribute feature vector to capture the influence of spatial features in different regions and the changing pattern of temporal features in the time dimension, thereby generating more representative comprehensive features.

[0072] First, the spatial feature vector and attribute feature vector come from the feature extraction layer, which respectively describe the spatial distribution characteristics of the data points and the dynamic activity of the time dimension. In the feature fusion process, the bidirectional gated recurrent unit network is used as the core processing unit, which can effectively capture forward and backward context information in time series data processing. For example, for a certain time point, the bidirectional GRU not only considers the feature influence of the previous time point, but also combines the features of the later time point, making the modeling of the current time point more comprehensive.

[0073] The introduction of the dynamic context-aware mechanism makes the fusion process more flexible and adaptive. Specifically, the dynamic context-aware mechanism dynamically adjusts the weight of the time series features according to the importance information of each region in the spatial feature vector. The importance of a region can be determined by analyzing the density information or spatial correlation value in the spatial feature vector. For example, assuming that the spatial feature vector of a region indicates that it contains multiple highly sensitive facilities (such as substations or airports), the importance of the region is given a higher weight, while regions with lower density or fewer facilities are given lower weights. This weight information is introduced into the weighted calculation of the time series feature vector through the dynamic context-aware mechanism, ensuring that regions with high importance contribute more to the results in the time series feature processing.

[0074] To illustrate with a specific example, assume that the data points have the following spatial feature weights in three regions: Region A has a weight of 0.6, Region B has a weight of 0.3, and Region C has a weight of 0.1. At a certain point in time, the time series feature vector contains the activity level, historical trends, and related events at that point in time. The dynamic context-aware mechanism weights these time series features according to the spatial weights. For example, in Region A, the activity level feature at that point in time will be amplified, while in Region C, the impact of this feature on the final comprehensive feature will be weakened accordingly. This weighted adjustment based on regional importance ensures that the model can focus on regions and time points that are more critical to sensitivity judgment.

[0075] The temporal features processed by the dynamic context-aware mechanism are input into the bidirectional GRU network and participate in fusion modeling together with the spatial features. The bidirectional GRU filters the input weighted temporal features through its gating unit, retains the key features, and suppresses noise or irrelevant information. Through multiple cycles, the network gradually forms a global comprehensive feature at each time point, which not only contains dynamic information in the time dimension, but also combines the regional characteristics of the spatial dimension.

[0076] Finally, the comprehensive feature vector output by the bidirectional GRU network reflects the dynamic changes of data points in the time dimension and captures their distribution pattern in the spatial dimension, providing high-quality input data for subsequent sensitivity assessment. This feature fusion method combined with dynamic context-aware mechanism can effectively improve the adaptability and recognition ability of sensitive data detection models for complex spatiotemporal data.

[0077] Furthermore, the dynamic context-aware mechanism of the bidirectional gated recurrent unit network used in the feature fusion layer is implemented by the following steps: Using the following formula (8), according to the spatial eigenvector The specific area characteristic value in Calculate the regional importance weight for each spatial region : ; in, Represents the total number of spatial regions; Represents a spatial region The normalized weight of is used to quantify the importance of the region; According to the following formula (9), the attribute feature of each time point in the attribute feature vector Calculating dynamic spatiotemporal weights : ; in is the time attenuation coefficient, For space area With time point Contextual relevance; According to the following formula (10), a comprehensive feature representation is generated : ; in Indicates the total number of time points; It is a balance parameter used to adjust the contribution ratio of spatial features and attribute features in the comprehensive feature vector; is the spatial feature vector.

[0078] First, according to the spatial eigenvector The regional characteristic values ​​in , calculate the regional importance weight of each spatial region . Formula (8) is defined as follows: ; in, Represents the region in the spatial feature vector Specific attribute values, such as region An indicator of the number of sensitive facilities or the importance of an area. represents the total number of spatial regions, It is a region The normalized weight of is used to quantify the importance of the region. Through the normalization operation, it is ensured that the sum of the importance weights of all regions is 1. For example, assuming that the spatial feature vector , the eigenvalues ​​corresponding to the three regions are and 30, then the regional weight is calculated as This means that region 2 contributes most to the comprehensive characteristics.

[0079] Next, the dynamic spatiotemporal weight of each time point in the attribute feature vector is calculated according to formula (9): The formula is defined as follows: ; in, It's time point The spatiotemporal weights of and Region With time point Contextual relevance Contextual relevance Can be defined by actual application scenarios, such as representing areas At the point in time The frequency of events or the activity of the space. is the time decay coefficient, which is used to control the influence of time interval on weight. The value will make the time difference more significant, and is usually in the range of 0.1 to 1.0. For example, assuming the area weight , context relevance , and , then for the time point , the dynamic weight is calculated as: ; The calculation result is .

[0080] Finally, based on the dynamic spatiotemporal weights and attribute feature vector , combined with the spatial eigenvector , generating a comprehensive feature representation . Formula (10) is defined as follows: ; in, Indicates the total number of time points, is the time point in the attribute feature vector The characteristic values ​​of Dynamic activity or other related information. Parameters is the balance coefficient, which is used to adjust the weights of spatial features and attribute features in the comprehensive features. The value of can be set according to specific application requirements, usually between 0.5 and 1.0. For example, for the attribute characteristics at three time points , dynamic spatiotemporal weight , spatial characteristics , balance coefficient , the comprehensive feature representation is calculated as: ; The final comprehensive feature representation combines time point features and spatial region features to provide a comprehensive and accurate representation of sensitive data.

[0081] Through the above steps, the dynamic context-aware mechanism can flexibly combine spatial and temporal information to generate a highly adaptable comprehensive feature representation, providing strong support for subsequent sensitivity analysis.

[0082] Step S104: input the target data set into the sensitive data detection model to obtain a model output result; based on the model output result, annotate the sensitive data included in the target data set to generate a sensitivity classification label; wherein the target data set uses preprocessed target structured data.

[0083] In step S104, the target data set is input into the trained sensitive data detection model. The model generates a corresponding model output result according to the feature vector of the input data, and further annotates the sensitive data contained in the target data set to generate a sensitivity classification label.

[0084] First, the target dataset is preprocessed to form input features that are consistent with the structured data format used when training the model. These features are represented as multidimensional feature vectors, which contain feature dimensions such as location distribution, spatial correlation, and data frequency to ensure that they fully match the input format of the model. The target dataset is input into the trained exploration model one by one or in batches, and the model uses the previously learned parameter weights to analyze and infer the feature vector of each record.

[0085] The model inference process is based on the forward propagation mechanism, that is, the input feature vector and the model weight are operated through each layer of the model structure in turn to calculate the probability distribution of each sensitivity category. For each input data, the model outputs a probability distribution vector, in which each element represents the probability that the data belongs to the corresponding sensitivity category. For example, possible classification labels include "high sensitivity", "medium sensitivity" and "low sensitivity", and the probability vector output by the model indicates the probability of the data in these categories.

[0086] Next, based on the probability distribution of the model output, the category with the highest probability value is selected as the preliminary sensitivity classification label of the data. To improve the reliability of the classification label, business rules or dynamic confidence thresholds can be combined to further judge certain boundary conditions. For example, when the probabilities of multiple classifications are close, they can be further weighted and adjusted according to specific feature dimensions (such as spatial correlation) to ensure that the classification results are more in line with actual application needs.

[0087] After the classification is completed, the generated sensitivity classification labels are attached to the target data set, and the classification results of each data are recorded in a structured manner. These classification labels can be stored in the database in the form of additional fields, corresponding one-to-one with the original data, thus forming a complete annotated data set. The sensitivity classification labels finally generated provide a clear basis for subsequent hierarchical protection and data management, and ensure that the exploration process fully covers and accurately annotates the target data set.

[0088] Step S104 uses the inference capability of the model to accurately associate each record in the target data set with the corresponding sensitivity classification label, laying the foundation for subsequent processing of sensitive data. The entire process is very efficient.

[0089] Step S105: Based on the sensitivity classification labels, the detected sensitive data is protected in a hierarchical manner, including encryption storage and access control strategies for highly sensitive data.

[0090] In step S105, the detected sensitive data is protected in a hierarchical manner according to the sensitivity classification label generated in step S104 to ensure data security and controllability of access, especially implementing strict encryption storage and access control strategies for highly sensitive data.

[0091] First, all data is classified into different protection levels according to sensitivity classification labels, including highly sensitive data, medium sensitive data, and low sensitive data. The protection measures for each category are gradually strengthened according to its sensitivity level to ensure the security and compliance of sensitive data. For highly sensitive data, strong encryption storage measures are adopted, such as using the Advanced Encryption Standard (AES) algorithm to store data in ciphertext form in a database or distributed storage system. The encryption process uses dynamically generated keys, and each access must be verified by the key to reduce the risk of key leakage. The generation and management of keys can be achieved through a secure key management system (KMS) to ensure the secure storage and access of keys.

[0092] After storage is completed, strict access control policies are set for access to highly sensitive data. Specifically, a multi-factor authentication mechanism is used to ensure that the identity of the visitor is credible, such as through a combination of passwords, fingerprints, or one-time verification codes. At the same time, role-based access control is applied to limit the access rights of different user roles, allowing only authorized users to access data related to their responsibilities. The allocation of access rights can be dynamically adjusted through the permission management module to adapt to changes in user roles or responsibilities.

[0093] For medium-sensitive data, lightweight encryption methods are used, such as simplified algorithms based on symmetric encryption, to ensure secure data storage. At the same time, simple access control based on IP addresses or regional restrictions is used to reduce unauthorized external access. Access rights to medium-sensitive data are mainly for internal authorized users, and access logs are retained to monitor abnormal access behavior.

[0094] For low-sensitivity data, encrypted storage is usually not required, but access logging is required. Each data access records the visitor's identity, access time, and operation type for subsequent audit and analysis. Although low-sensitivity data does not directly involve high risks, access logging can provide basic security and detect potential risk behaviors.

[0095] Finally, the accuracy of data classification and the effectiveness of protection measures can be ensured by regularly auditing and optimizing the hierarchical protection mechanism. The audit process can evaluate data storage, encryption strength, access behavior, and permission settings based on automated tools, and dynamically adjust protection measures based on security policy updates and the emergence of new threats. Through this hierarchical protection mechanism, the detected sensitive data can be fully and effectively protected, the security of highly sensitive data can be ensured, and resource utilization and management efficiency can be optimized.

[0096] In the above-mentioned embodiment, a sensitive data detection method based on machine learning is provided. Correspondingly, the present application also provides a sensitive data detection system based on machine learning. Figure 2, which is a schematic diagram of an embodiment of a sensitive data detection system based on machine learning of the present application. Since this embodiment, i.e., the second embodiment, is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The system embodiment described below is only illustrative.

[0097] A second embodiment of the present application provides a sensitive data detection system based on machine learning, including: The collection unit 201 is used to collect raw data including tower location data, airport location data, and important station location data from power-related data sources; pre-process the raw data, wherein the formatting process includes cleaning abnormal values, unifying data formats, and standardizing geographic coordinates to generate pre-processed structured data; A generating unit 202 is used to generate a multidimensional feature vector based on the preprocessed structured data; according to a predetermined sensitivity labeling rule, the multidimensional feature vector is combined with its corresponding sensitivity labeling label to generate model training data; wherein the multidimensional feature vector includes position distribution, spatial correlation, and data frequency; A construction unit 203 is used to construct a sensitive data detection model based on the model training data using a machine learning algorithm; train the sensitive data detection model through iterative optimization to obtain model parameters for identifying the sensitivity of target data; and obtain a trained sensitive data detection model according to the model parameters; The labeling unit 204 is used to input the target data set into the sensitive data detection model to obtain a model output result; according to the model output result, label the sensitive data included in the target data set to generate a sensitivity classification label; wherein the target data set uses the preprocessed target structured data; The grading unit 205 is used to perform graded protection on the detected sensitive data according to the sensitivity classification labels, including encryption storage and access control strategies for highly sensitive data.

[0098] A third embodiment of the present application provides an electronic device, the electronic device comprising: processor; The memory is used to store a program, and when the program is read and executed by the processor, it executes a sensitive data detection method based on machine learning provided in the first embodiment of the present application.

[0099] The fourth embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, a sensitive data detection method based on machine learning provided in the first embodiment of the present application is executed.

[0100] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

Claims

1. A sensitive data detection method based on machine learning, characterized in that: include: Collect raw data including tower location data, airport location data, and important station location data from power-related data sources; Preprocessing the raw data, wherein the preprocessing includes cleaning outliers, unifying data formats and standardizing geographic coordinates to generate preprocessed structured data; Based on the preprocessed structured data, a multidimensional feature vector is generated; according to a predetermined sensitivity labeling rule, the multidimensional feature vector is combined with its corresponding sensitivity labeling label to generate model training data; wherein the multidimensional feature vector includes location distribution, spatial correlation, and data frequency; Based on the model training data, a sensitive data detection model is constructed using a machine learning algorithm; the sensitive data detection model is trained by iterative optimization to obtain model parameters for identifying the sensitivity of target data; and a trained sensitive data detection model is obtained according to the model parameters; Input the target data set into the sensitive data detection model to obtain the model output result; according to the model output result, label the sensitive data included in the target data set to generate a sensitivity classification label; wherein the target data set uses the preprocessed target structured data; Based on the sensitivity classification labels, the detected sensitive data is protected in different levels, including encrypted storage and access control strategies for highly sensitive data.

2. The sensitive data detection method based on machine learning according to claim 1, characterized in that: The method of generating a multi-dimensional feature vector based on the pre-processed structured data includes: Divide the input geographic data into equidistant grid cells according to a preset rule, wherein each grid cell corresponds to a specified geographic range; calculate the number of geographic data points in each grid cell, expressed as grid density; normalize the grid density value to generate a location distribution feature reflecting the local distribution feature of the geographic data; For each geographic data point, calculate its distance from a set of predefined important facilities; assign weights according to facility types, perform weighted average of each distance value and the corresponding weight to obtain a spatial correlation value; normalize the spatial correlation values ​​of all geographic data points to generate spatial correlation features; wherein the important facilities include towers, airports and substations; The time attributes of each geographic data point are analyzed, and the number of occurrences within a fixed time period is counted as the frequency value; by setting the time decay weight, the frequency value is weighted and normalized to generate the spatiotemporal frequency characteristics that describe the spatiotemporal activity level; The normalized location distribution features, spatial correlation features, and spatiotemporal frequency features of each geographic data point are combined to generate a multidimensional feature vector, and a feature vector dataset is output for model training.

3. The sensitive data detection method based on machine learning according to claim 2 is characterized in that: For each geographic data point, the distance between the point and a set of predefined important facilities is calculated; weights are assigned according to facility types, and each distance value is weighted averaged with the corresponding weight to obtain a spatial association value; The spatial correlation values ​​of all geographic data points are normalized to generate spatial correlation features, including: For each geographic data point , calculate its important facilities set Each facility distance, generate a distance matrix , among which, Line Column value Indicates geographic data points to The distance to the facility, Calculate according to the following formula (1): ； in, and Respectively geographic data points and The location coordinates of each facility; According to Risk factor of important facilities , a dynamic weight is assigned to each facility according to the following formula (2): ； in, For the The weight coefficient of each important facility; The number of important facilities; According to the following formula (3), the spatial association value of the geographic data point is calculated : ； in, is the attenuation coefficient, which is used to control the effect of distance on spatial correlation; For the matrix No. The maximum distance value of the column, used for normalized distance; According to the following formula (4), the spatial association value of all geographic data points is Perform normalization: ； in, Normalized spatial correlation value; For all spatial association values The minimum value in ; For all spatial association values The maximum value in ; A spatial correlation feature is generated according to the normalized spatial correlation value.

4. The sensitive data detection method based on machine learning according to claim 2 is characterized in that: The time attribute of each geographic data point is analyzed, and the number of occurrences within a fixed time period is counted as a frequency value; by setting a time decay weight, the frequency value is weighted and normalized to generate a spatiotemporal frequency feature describing the spatiotemporal activity, including: For geographic data points at fixed time periods The number of occurrences within , according to the decay weight at the current time point Weighted, where the attenuation weight Determined by the following formula (5): ； in, Indicates the current time and time period time interval; is the attenuation control parameter, which is used to adjust the weight impact of long-term data; It is an adjustable exponential parameter used to control the nonlinear degree of attenuation; In each time period According to the following formula (6), the weighted frequency value is calculated: : ； in, is the total number of time periods; is the weighted cumulative frequency value of the geographic data point in all time periods; According to the following formula (7), the normalized spatiotemporal frequency feature is generated : ； in, is the weighted frequency mean of all time periods; is the standard deviation of the weighted frequency values, which is used to measure the difference in activity of a data point relative to the overall distribution.

5. The sensitive data detection method based on machine learning according to claim 1, characterized in that: The sensitive data detection model includes a feature extraction layer, a feature fusion layer and a sensitivity assessment layer; The feature extraction layer is used to receive the pre-processed target structured data; the feature extraction layer uses a parallel spatial feature extraction branch and an attribute feature extraction branch to process the received data respectively; the spatial feature extraction branch is implemented using a spatial convolutional network, which is used to extract the position distribution characteristics and spatial correlation characteristics of each data point through multi-layer convolution to generate a spatial feature vector; the attribute feature extraction branch is implemented using an attention mechanism to capture the temporal correlation characteristics and activity level of the data point to generate an attribute feature vector; The feature fusion layer is used to receive the spatial feature vector and the attribute feature vector provided by the feature extraction layer; the feature fusion layer is implemented using a bidirectional gated recurrent unit network to process the received data and generate a unified comprehensive feature vector; The sensitivity assessment layer is used to receive the comprehensive feature vector provided by the feature fusion layer; the sensitivity assessment layer is used to input the comprehensive feature vector into a multi-layer fully connected network to calculate the sensitivity score of each data point; the scoring result is calibrated according to predefined sensitivity rules; based on the sensitivity score and rule calibration results, a dynamic confidence threshold is used to determine whether the data point belongs to a specific sensitivity category, and a sensitivity classification label is generated.

6. The sensitive data detection method based on machine learning according to claim 5, characterized in that: The spatial convolutional network used by the spatial feature extraction branch includes an input layer, a local receptive field convolutional layer, a cross-region convolutional layer, a multi-scale feature fusion layer, a normalization and output layer; The input layer is used to receive the preprocessed target structured data, organize the data containing the geographical location information and related regional features of each data point into a two-dimensional matrix form; each row of the two-dimensional matrix represents the spatial information of a data point, and each column represents the specific features of the data point, thereby generating an input feature matrix; The local receptive field convolution layer is used to process the input feature matrix, and by setting a convolution kernel with a fixed range, performs a convolution operation on the features of each data point and its neighboring data points, extracts the local spatial distribution characteristics, and generates a preliminary local feature map, which is used to describe the neighborhood distribution pattern of the data points and the feature distribution density of the local area; The cross-region convolution layer is used to receive the local feature map and extract the potential spatial correlation between data points in a large range by expanding the convolution operation of the receptive field range; using multiple convolution kernels of different sizes, cross-region features are extracted at different receptive field scales to generate a global spatial feature map to reflect the correlation and spatial characteristics between distant data points; The multi-scale feature fusion layer is used to receive the local feature map and the global spatial feature map, and fuse the two according to a specified weight ratio; the fused comprehensive feature map provides a unified description of the local and global relationship of the data points; The normalization and output layer is used to perform batch normalization on the comprehensive feature map, balance the numerical differences between the output features of different convolutional layers, and generate a normalized spatial feature vector.

7. The sensitive data detection method based on machine learning according to claim 5, characterized in that: The attention mechanism of the attribute feature extraction branch includes the following implementation process: receiving the preprocessed target structured data, and representing the time attribute as a time series feature vector; Calculate the context weight for the feature vector at each time point and dynamically assign weights based on time differences and feature similarities; Use context weights to perform weighted aggregation on time series feature vectors to generate enhanced time feature vectors; All enhanced temporal feature vectors are integrated into a global temporal feature representation and output to the feature fusion layer for further processing.

8. The sensitive data detection method based on machine learning according to claim 5, characterized in that: The bidirectional gated recurrent unit network used in the feature fusion layer jointly models the correlation and spatial distribution pattern of the spatial feature vector and the attribute feature vector in the time dimension by introducing a dynamic context-aware mechanism, wherein the dynamic context-aware mechanism dynamically adjusts the weight of the temporal feature according to the regional importance of the spatial feature.

9. The sensitive data detection method based on machine learning according to claim 8, characterized in that: The dynamic context-aware mechanism of the bidirectional gated recurrent unit network used in the feature fusion layer is implemented by the following steps: Using the following formula (8), according to the spatial eigenvector The specific area characteristic value in Calculate the regional importance weight for each spatial region : ； in, Represents the total number of spatial regions; Represents a spatial region The normalized weight of is used to quantify the importance of the region; According to the following formula (9), the attribute feature of each time point in the attribute feature vector Calculating dynamic spatiotemporal weights : ； in is the time attenuation coefficient, For space area With time point Contextual relevance; According to the following formula (10), a comprehensive feature representation is generated : ； in Indicates the total number of time points; It is a balance parameter used to adjust the contribution ratio of spatial features and attribute features in the comprehensive feature vector; is the spatial feature vector.

10. A sensitive data detection system based on machine learning, characterized in that: include: A collection unit, used to collect raw data including tower location data, airport location data, and important station location data from power-related data sources; Preprocessing the raw data, wherein the preprocessing includes cleaning outliers, unifying data formats and standardizing geographic coordinates to generate preprocessed structured data; A generating unit, configured to generate a multidimensional feature vector based on the preprocessed structured data; and to combine the multidimensional feature vector with its corresponding sensitivity label according to a predetermined sensitivity labeling rule to generate model training data; wherein the multidimensional feature vector includes position distribution, spatial correlation, and data frequency; A construction unit is used to construct a sensitive data detection model based on the model training data using a machine learning algorithm; train the sensitive data detection model through iterative optimization to obtain model parameters for identifying the sensitivity of target data; and obtain a trained sensitive data detection model according to the model parameters; A labeling unit, used for inputting a target data set into the sensitive data detection model to obtain a model output result; labeling the sensitive data included in the target data set according to the model output result to generate a sensitivity classification label; wherein the target data set adopts the preprocessed target structured data; The grading unit is used to perform graded protection on the detected sensitive data according to the sensitivity classification labels, including encryption storage and access control strategies for highly sensitive data.

Images