Natural resource data security management method

Abstract

The invention discloses a natural resource data security management method, and relates to the field of natural resource data security management, and the method comprises the steps: obtaining a target field set requested to be accessed by a user and knowledge graph data, traversing fields, and querying graph association nodes to generate an extended field set; then calculating a conduction risk value of a field pair formed by any two fields in the extension set, if a risk mean value of all the field pairs reaches an evolution threshold value, screening out a high-risk field pair, extracting common associated nodes connected with the high-risk field pair in the atlas, injecting the common associated nodes into the extension set, and carrying out loop iteration on the process until a termination condition is met; and finally, calculating a global risk entropy based on the risk values and sensitivity interference factors of all fields in the extension set, performing desensitization operation on the target field set, and synchronously updating the knowledge graph database. According to the method, the implicit sensitive information conduction risk in the natural resource data multi-field combination query can be dynamically identified and quantified.

Description

Technical Field

[0001] This invention relates to the field of natural resource data security management, specifically a method for natural resource data security management. Background Technology

[0002] Natural resource data is crucial foundational data for land spatial governance, ecological environmental protection, and resource development and management, encompassing multi-dimensional information such as geographic coordinates, land type and area, and geological structure. With the development of digitalization and informatization, the scenarios for the collection, sharing, and application of natural resource data are constantly expanding, and government agencies, research institutions, and enterprises are increasingly demanding cross-departmental and cross-system data fusion and intelligent analysis. In this process, data security management has become a critical link in ensuring the compliant use of natural resource data and preventing its leakage.

[0003] Traditional data security control mechanisms primarily rely on static sensitivity classification and access permission policies to control access or anonymize single fields or fixed datasets. However, this approach cannot meet the needs of complex queries, dynamic access, and cross-system combined data analysis. In existing technologies, the security protection of natural resource data is typically based on static sensitive field labels or anonymization strategies triggered by a single data request. For example, by pre-setting labels such as "geographic coordinates = high risk," sensitive fields are anonymized in a fixed manner during the data output stage.

[0004] However, such static strategies have shortcomings: they cannot effectively deal with the risk of implicit sensitive information deduction during multi-field joint queries and multiple accesses. In particular, when users deduce cross-field information through knowledge graph relationships, traditional methods are difficult to identify and block such potential combined risks, and are also difficult to prevent the leakage of sensitive information caused by the accumulation of long-term access behavior.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method for the secure management of natural resource data.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] In a first aspect, the present invention discloses a method for the security management of natural resource data, comprising the following steps:

[0009] Obtain the target field set currently requested by the user and the knowledge graph database data; the target field set is at least one of geographic coordinates, land area, and geological structure.

[0010] Iterate through each field in the target field set, query the set of related nodes in the knowledge graph library, and generate an extended field set;

[0011] The transmission risk value of a field pair is calculated by weighting any two different fields from the extended field set.

[0012] Determine whether the mean of the transmission risk value of all fields reaches the preset evolution threshold. If so, select the field pairs whose transmission risk value is greater than the preset risk threshold as high-association field pairs. Extract the common association node set that is connected to the high-association field pairs from the knowledge graph library. Inject the common association node set into the extended field set and re-determine whether the evolution threshold is reached until the preset termination condition is met.

[0013] Calculate the global risk entropy based on the transmission risk values ​​of all field pairs in the extended field set and the preset sensitivity interference factor;

[0014] The target field set is anonymized based on the global risk entropy, and the knowledge graph database data is updated.

[0015] Secondly, this invention discloses a natural resource data security management system, comprising:

[0016] The data acquisition module is used to acquire the target field set currently requested by the user and the knowledge graph database data; the target field set includes at least one of geographic coordinates, land area, and geological structure;

[0017] The field expansion generation module is used to traverse each field in the target field set, query the corresponding set of related nodes based on the knowledge graph library, and generate an expanded field set based on each set of related nodes.

[0018] The transmission risk calculation module is used to form field pairs from any two different fields in the extended field set, and to perform weighted calculations on each field pair to obtain the corresponding transmission risk value.

[0019] The high-association node injection module is used to determine whether the average value of the transmission risk of all fields reaches a preset evolution threshold. If so, the field pairs with transmission risk values ​​greater than the preset risk threshold are selected as high-association field pairs. The module extracts a set of common association nodes that are connected to the high-association field pairs from the knowledge graph library, injects the set of common association nodes into the extended field set, and re-determines whether the evolution threshold is reached until the preset termination condition is met.

[0020] The global risk entropy calculation module is used to calculate the global risk entropy based on the transmission risk values ​​of all field pairs in the extended field set and the preset sensitivity interference factor;

[0021] The desensitization processing module is used to perform desensitization operations on the target field set according to the global risk entropy, and update the knowledge graph database data based on the desensitization operation results.

[0022] Compared with existing technologies, the advantages of this invention are as follows: it can dynamically identify and quantify the risk of implicit sensitive information transmission in multi-field combined queries of natural resource data. By generating and iteratively updating the extended field set, the system captures potential high-risk transmission paths in the knowledge graph. The calculation of the weighted transmission risk value takes into account the time dimension, effectively addressing the risks caused by the accumulation of long-term access behavior. The introduction of global risk entropy provides a quantitative triggering mechanism for desensitization operations, realizing dynamic monitoring and risk assessment of the entire data access process. This method significantly improves the system's security protection capabilities for complex query scenarios, effectively preventing the risk of sensitive information being gradually deduced through multiple legitimate queries, and providing a more flexible and precise protection mechanism for the secure management of natural resource data. Attached Figure Description

[0023] 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is an overall block diagram of the method in Embodiment 1 of the present invention;

[0025] Figure 2 This is a flowchart of the method according to Embodiment 1 of the present invention;

[0026] Figure 3 This is an overall block diagram of the system in Embodiment 2 of the present invention. Detailed Implementation

[0027] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0028] Application Overview: In traditional natural resource data security management mechanisms, static sensitivity classification and fixed access permission strategies are insufficient to address the implicit inference risks in multi-field joint query scenarios. Due to the lack of dynamic expansion analysis of related nodes in the knowledge graph, the system cannot identify potential sensitive information transmission paths formed by cross-field combinations. When users submit combined query requests containing geographic coordinates and land area, traditional methods only anonymize the original fields, ignoring the indirect relationships formed through geological structure nodes in the knowledge graph, resulting in the transmission risks between sensitive fields not being quantitatively assessed.

[0029] For example, in a cross-regional mineral resource assessment scenario, a user requests the geographical coordinates of a mining area and the area of ​​surrounding land types via an API interface. Traditional systems add Gaussian noise to the geographical coordinates based on preset labels, but fail to detect any correlation between these geographical coordinates and groundwater level monitoring nodes through a knowledge graph. When the same user subsequently requests geological structure data for the same area, the system does not establish a correlation analysis between the two queries, and cannot prevent the implicit correlation path between the precise coordinates of the mining area and mineral reserves from being deduced through time-series queries.

[0030] If the aforementioned issues are not addressed, the risk of data transmission between heterogeneous fields from multiple sources will lead to the gradual deduction of sensitive information through multiple legitimate queries, forming a hidden channel for data leakage. The system's inability to dynamically identify highly correlated field pairs and their common neighbor nodes results in the continuous accumulation of high-risk paths in the knowledge graph. In the long term, this will trigger security vulnerabilities during cross-system data fusion, allowing attackers to reverse-engineer high-value information by combining low-sensitivity fields, ultimately leading to the unauthorized leakage of core natural resource data in compliant usage scenarios.

[0031] Example 1:

[0032] like Figure 1-2As shown, a method for managing the security of natural resource data includes the following steps: obtaining the target field set currently requested by the user and the knowledge graph database data; the target field set is at least one of geographic coordinates, land area, and geological structure; traversing each field in the target field set, querying the set of related nodes in the knowledge graph database, and generating an extended field set; forming a field pair from any two different fields in the extended field set, and calculating the transmission risk value of the field pair by weighting; determining whether the mean of the transmission risk values ​​of all field pairs reaches a preset evolution threshold, and if so, selecting field pairs with transmission risk values ​​greater than the preset risk threshold as highly correlated field pairs, extracting a set of common related nodes connected to the highly correlated field pairs from the knowledge graph database, injecting the set of common related nodes into the extended field set, and re-determining whether the evolution threshold is reached, until a preset termination condition is met; calculating the global risk entropy based on the transmission risk values ​​of all field pairs in the extended field set and a preset sensitivity interference factor; performing a desensitization operation on the target field set according to the global risk entropy, and updating the knowledge graph database data.

[0033] The target field set refers to the natural resource data fields requested by the user, including at least one of geographic coordinates, land area, and geological structure. This is achieved by parsing field identifiers in the user's query, limiting the initial access scope and serving as the starting point for subsequent risk analysis. The knowledge graph database is a structured database storing the relationships between natural resource data, specifically implemented using a graph database to store field nodes and their associated edges, providing a foundation for semantic relationships and derivation path analysis between fields. The extended field set is an extended dataset generated by traversing the associated nodes of the target field in the knowledge graph, specifically implemented by using a breadth-first search algorithm to query and merge the sets of one-hop neighbor nodes, used to discover hidden risk fields with potential derivation relationships to the target field. The transmission risk value is an indicator that quantifies the probability of sensitive information derivation between field pairs, specifically implemented by a weighted calculation of a normalized derivation risk coefficient and a time density decay factor, used to dynamically assess the risk of implicit information leakage caused by multi-field combined queries. The evolution threshold is the risk threshold that triggers iterative updates to the extended field set, specifically implemented by comparing the average of the transmission risk values ​​of the statistical fields with a preset threshold, used to control the dynamic expansion boundary of the risk transmission path. Highly correlated field pairs refer to field combinations whose transmitted risk values ​​exceed a preset risk threshold. This is achieved by using a sorting and filtering algorithm to select the top N% of field pairs by risk value, used to identify high-risk derivation paths and locate key correlated nodes. A set of commonly correlated nodes refers to the set of nodes that are directly connected to all highly correlated field pairs. This is achieved by using a bidirectional traversal algorithm to obtain the intersection of neighboring nodes, used to inject new risk transmission paths to improve the global risk assessment model. Global risk entropy refers to the overall risk index that combines transmitted risk values ​​and sensitivity interference factors. This is achieved by using a linear weighted summation formula to calculate the risk contribution value of all field pairs, used to quantify the global sensitive information exposure degree of the target field set. Desensitization operation refers to data transformation processing of the target fields based on global risk entropy, specifically using Gaussian noise addition or numerical scaling algorithms, used to block sensitive information derivation paths while ensuring data availability.

[0034] The core innovation of this application lies in dynamically expanding the field set through the knowledge graph, combining iterative calculation of the transmitted risk value with the injection mechanism of common related nodes, to construct a global risk assessment model covering multi-hop derivation paths, thereby realizing dynamic desensitization decision-making and real-time updating of the knowledge graph in complex query scenarios for natural resource data.

[0035] The working process and principle of this application are as follows: First, the target field set requested by the user and the knowledge graph database data are obtained. The target field set includes at least one of geographic coordinates, land area, and geological structure. Then, each field in the target field set is traversed, and the set of related nodes in the knowledge graph database is queried to generate an extended field set. Any two different fields in the extended field set are used to form a field pair, and the transmission risk value of the field pair is calculated by weighting. Next, it is determined whether the mean of the transmission risk values ​​of all field pairs reaches a preset evolution threshold. If it does, field pairs with transmission risk values ​​greater than the preset risk threshold are selected as highly correlated field pairs. A set of common related nodes connected to the highly correlated field pairs is extracted from the knowledge graph database, and the set of common related nodes is injected into the extended field set. The evolution threshold is re-evaluated until a preset termination condition is met. Based on the transmission risk values ​​of all field pairs in the extended field set and the preset sensitivity interference factor, the global risk entropy is calculated. Finally, the target field set is desensitized according to the global risk entropy, and the knowledge graph database data is updated.

[0036] This solution identifies the risk of implicit sensitive information transmission in multi-field combined queries by dynamically expanding the field set and iteratively injecting highly correlated nodes. The calculation of the weighted transmission risk value takes into account the time dimension, capturing the risks caused by the accumulation of long-term access behavior. The introduction of global risk entropy provides a quantitative indicator for triggering de-identification operations. The entire process forms a closed-loop risk assessment and protection mechanism, effectively addressing data security challenges in complex query scenarios.

[0037] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0038] In the natural resources data management system, users request access to the geographic coordinates and land area data of a mining area via an API interface. The system first obtains these two target fields and then reads data from a pre-built knowledge graph database. The knowledge graph database contains the relationships between various fields of natural resources.

[0039] The system iterates through the geographic coordinates and land area fields, querying the knowledge graph for nodes directly connected to them. For example, geographic coordinates might be connected to nodes such as topographic elevation and administrative division, while land area might be connected to nodes such as land use type and vegetation cover. These associated nodes are added to the extended field set.

[0040] Next, the system constructs field pairs from any two different fields in the extended field set. For each field pair, the system extracts a pre-calculated derivation risk coefficient from the knowledge graph database and calculates a time density decay factor based on recent query logs. These two factors are then weighted and combined to obtain the transmission risk value of the field pair.

[0041] The system calculates the mean of the transmission risk value for all fields and compares it with a preset evolution threshold. If the mean exceeds the threshold, the system will filter out field pairs with transmission risk values ​​higher than the preset threshold, such as (geographic coordinates, terrain elevation). Then, it searches the knowledge graph for nodes connected to both fields, such as geological structures, and injects them as common associated nodes into the extended field set.

[0042] The system repeats the above process until the preset termination conditions are met, such as the number of iterations reaching the upper limit or the overlap between the newly injected node and the existing set being too high.

[0043] Finally, the system calculates the global risk entropy based on the transmission risk values ​​of all field pairs in the final extended field set, combined with preset sensitivity interference factors (such as the factor weight of geographic coordinates being higher than that of land area).

[0044] If the global risk entropy exceeds the preset desensitization threshold, the system performs desensitization operations on the target fields. For example, Gaussian noise is added to geographic coordinates, and the area of ​​land categories is scaled proportionally. At the same time, the system updates the derived risk coefficients of relevant field pairs in the knowledge graph database to reflect the impact of this visit on future risk assessments.

[0045] Through the above scheme, this application can dynamically identify and quantify the risk of implicit sensitive information transmission in multi-field combined queries of natural resource data. By generating and iteratively updating the extended field set, the system captures potential high-risk transmission paths in the knowledge graph. The calculation of the weighted transmission risk value takes into account the time dimension, effectively addressing the risks caused by the accumulation of long-term access behavior. The introduction of global risk entropy provides a quantitative triggering mechanism for desensitization operations, realizing dynamic monitoring and risk assessment of the entire data access process. This method significantly improves the system's security protection capabilities for complex query scenarios, effectively preventing the risk of sensitive information being gradually deduced through multiple legitimate queries, and providing a more flexible and precise protection mechanism for the secure management of natural resource data.

[0046] In some of the solutions described above in this application, if only the fields of the target field set itself are considered when generating the extended field set, it may result in insufficient coverage of associated nodes, failure to fully identify potential risk nodes, and affect the comprehensiveness of subsequent risk value calculation.

[0047] This application further proposes to traverse each field in the target field set, query the knowledge graph database for a set of 1-hop neighbor nodes that have a direct relationship with that field, and then take the union of the target field set and all the queried 1-hop neighbor node sets to generate an initial extended field set.

[0048] When traversing each field in the target field set, the node for each field in the knowledge graph is used as the starting point for the query operation. The one-hop neighbor node set is defined as the set of nodes directly connected to the starting node by a single edge; the edge type includes attribute association or semantic association. The union operation combines the original target field set with the neighbor node sets corresponding to each field using set operations, eliminating duplicate nodes to form the initial extended field set. For example, when the target field set contains a geographic coordinate field, its one-hop neighbors might include nodes such as elevation data and administrative division codes.

[0049] Specifically, in the knowledge graph database, each field node establishes relationships with other nodes through predefined edge connection rules. Query operations are implemented based on the graph database's adjacency query interface, performing a breadth-first search algorithm for each target field node, limiting the search depth to one level. The obtained set of neighboring nodes is then combined with the original target field set to ensure that the expanded field set includes both the original requested field and its directly associated nodes. In this way, the coverage of the initial expanded field set extends from the original requested field to its directly associated nodes, providing a more comprehensive data foundation for subsequent risk propagation calculations. For example, if the first-hop neighbor of the local soil type area field contains soil type distribution data, this data will be included in the expanded field set for risk calculation, thereby improving the ability to identify potential information leakage paths.

[0050] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0051] Iterate through each field in the target field set and query the knowledge graph for the set of one-hop neighbor nodes that have a direct relationship with that field. For example, for the target field "geographic coordinates", possible one-hop neighbor nodes include "administrative division" and "topographic features". For the target field "land type area", possible one-hop neighbor nodes include "land use type" and "ecological function zone".

[0052] The initial extended field set is generated by taking the union of the target field set and the set of all 1-hop neighbor nodes obtained from the query. Specifically, if the target field set is {"geographic coordinates", "land area"}, and the 1-hop neighbor node sets obtained from the query are {"administrative division", "topographic features"} and {"land use type", "ecological function zone"}, then the initial extended field set is {"geographic coordinates", "land area", "administrative division", "topographic features", "land use type", "ecological function zone"}.

[0053] Through the above technical solution, this application achieves an effective expansion of the target field set. As a result, the system can capture additional information directly related to the original requested fields, providing a more comprehensive data foundation for subsequent risk assessment. Furthermore, by merging the target field with its one-hop neighbor node set, a more complete field network is formed, which helps identify potential data correlations and sensitive information propagation paths. This expansion method not only improves the accuracy of risk assessment but also anticipates possible data combination risks, thereby achieving more proactive and comprehensive data security management.

[0054] In some of the solutions described above in this application, after generating the extended field set, it is necessary to calculate the transmission risk value of the field pairs. However, traditional methods assess risk based on only a single static factor, which cannot dynamically reflect the impact of real-time query behavior on the derivation of sensitive information, resulting in insufficient accuracy of risk assessment.

[0055] This application further proposes a calculation process for the transmission risk value, including: for the field pair (p, q), extracting and deriving the risk coefficient from the knowledge graph database. ; Calculate the time density decay factor based on real-time query logs ; Derivation of risk coefficient and time-density decay factor Normalization is performed, and the transmission risk value is calculated. , where α and β are preset proportional coefficients.

[0056] Among them, the risk coefficient is derived. By statistically analyzing the frequency of field pairs in historical data breach cases, a long-term data breach risk can be assessed; a time-density decay factor is also used. The standard deviation of query timestamps within the most recent T hours is used to capture risk fluctuations caused by sudden surges in access. After normalization of both coefficients, they are linearly combined using preset proportional coefficients α and β to form a comprehensive risk assessment model. The proportional coefficients α and β are set to range from 0.6 to 0.8 and from 0.2 to 0.4, respectively, to ensure that historical risk factors dominate while also considering the impact of real-time behavior.

[0057] Specifically, when calculating the transmission risk value of the geological structure and land area field pair in a certain access request, the frequency of this field pair in the sensitive derivation path is first extracted from the historical leakage case database. The frequency of this field pair is 85 times, and the highest frequency is 100 times. The normalized value is then obtained. The value was 0.85. Subsequently, the field was analyzed for six query timestamps within the most recent 24 hours, and the standard deviation σ was calculated to be 3.2 hours. Substituting this into the S-curve function yielded... The value is 0.72. When the preset values ​​are α=0.7 and β=0.3, the final transmission risk value is 0.811. This calculation method effectively identifies risk combinations that, although not reaching the historical high-frequency leakage, exhibit abnormal access behavior by integrating long-term statistical characteristics and short-term access patterns.

[0058] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0059] The calculation process of the transmission risk value includes: for the field pair (p, q), extracting and deriving the risk coefficient from the knowledge graph database. ; Calculate the time density decay factor based on real-time query logs ; Derivation of risk coefficient and time-density decay factor Normalization is performed, and the transmission risk value is calculated: Where α and β are preset proportionality coefficients.

[0060] In practice, α=0.6 and β=0.4 can be set. First, the derivation risk coefficient of the field pair (p,q) is extracted from the knowledge graph database. ,For example =0.8. Then, the time density decay factor is calculated based on the real-time query logs. Assuming =0.5. (Regarding...) and Normalization is performed to ensure that all values ​​fall within the range of [0,1]. Finally, the values ​​are substituted into the formula to calculate the transmission risk value: =0.60.8+0.40.5=0.68.

[0061] Through the above technical solution, this application achieves dynamic assessment of the risk transmission of field pairs. Therefore, the system can accurately quantify the risk level of different field combinations based on the derivation relationships in the knowledge graph and real-time query behavior, providing a basis for subsequent data security management. Furthermore, by introducing a time-density decay factor, this solution can capture the additional risks that may arise from frequent queries of certain field combinations within a short period, thereby more comprehensively assessing the security of data access behavior.

[0062] In some of the solutions mentioned above in this application, a method for calculating the transmission risk value was proposed. However, in dynamic access scenarios, relying solely on statically derived risk coefficients is insufficient to accurately reflect the temporal distribution characteristics of real-time query behavior. Furthermore, the statistical methods for historical leakage cases cannot effectively quantify the differences in derivation path risk between different field pairs, resulting in a lack of dynamic adaptability and historical experience support in the assessment of the transmission risk value.

[0063] This application further proposes a method for calculating the time density decay factor as follows: obtain the query timestamp sequence of the field pair within the most recent T hours, calculate the standard deviation σ of the timestamps, and then calculate the time density decay factor.

[0064]

[0065] Where γ and λ are preset adjustment parameters; the initial construction method for deriving the risk coefficient is as follows: sensitive derivation paths are extracted from the historical natural resource leakage case database, and each path contains a sequence of field pairs that lead to information leakage; for each field pair in the path, its frequency of occurrence in all paths is counted. Calculate the initial derived risk coefficient:

[0066]

[0067] in The highest frequency among all field pairs.

[0068] The calculation of the time density decay factor quantifies the dispersion of query time distribution through standard deviation. A larger standard deviation indicates lower temporal clustering of query behavior. The standard deviation is mapped to a decay factor value using the Sigmoid function. γ is used to adjust the amplitude range of the decay factor, and λ controls the sensitivity of the standard deviation to the decay factor. The construction of the risk coefficient is based on path analysis of historical leakage cases. Frequency statistics reflect the activity level of field pairs in sensitive derivation paths, and normalization ensures that the risk coefficient falls within the [0,1] interval.

[0069] Specifically, after obtaining the query timestamp sequence of field pairs, the standard deviation is calculated using statistical methods to measure the dispersion of the timestamp distribution. For example, when the standard deviation is below a threshold, the time density decay factor approaches the γ value, and vice versa. The non-linear characteristics of the Sigmoid function can convert the standard deviation into a smooth decay coefficient, avoiding the impact of abrupt changes on the transmission risk value. In the process of constructing the risk coefficient derivation, each sensitive derivation path in the historical case library is decomposed into a sequence of field pairs. High-frequency dangerous field combinations are identified through frequency statistics. For example, when the combination of geological structure and land area appears most frequently in historical cases, its derivation risk coefficient is set to 1. Normalization ensures that the risk coefficients of different field pairs are comparable. For example, when a field pair appears 50 times and the highest frequency is 100 times, its risk coefficient is 0.5. The linear weighted combination of the time density decay factor and the derived risk coefficient can simultaneously reflect the time characteristics of real-time query behavior and the statistical regularity of historical leakage cases. For example, when a certain field is intensively queried in a short period of time and has a high historical risk coefficient, its transmission risk value will be significantly increased, triggering subsequent security processing mechanisms.

[0070] Through the above technical solution, this application can dynamically calculate the transmission risk of field pairs based on historical leakage cases and real-time query behavior, effectively capturing potential security risks in multi-field combination queries. The introduction of the time density decay factor enables the system to identify abnormal behavior of frequently querying specific field combinations within a short period of time, improving the sensitivity to changes in data access patterns. Simultaneously, by analyzing the frequency of field pairs in historical leakage cases, the system can more accurately assess the potential risks of different field combinations, thus providing a more precise basis for subsequent risk control and data anonymization.

[0071] In some of the above-mentioned schemes in this application, after filtering highly correlated field pairs, it is necessary to extract the common associated node set that is connected to them by edges. However, when traditional methods traverse the knowledge graph database, the node set may not be fully covered or irrelevant nodes may be introduced due to unidirectional search or path redundancy, which will affect the accuracy of subsequent risk calculation.

[0072] This application further proposes a process for extracting the common associated node set, which includes: performing a bidirectional traversal in the knowledge graph database, and for the field pair (p,q), obtaining the 1-hop neighbor node set P starting from p and the 1-hop neighbor node set Q starting from q; the common associated node set is the intersection of the neighbor node set P and the neighbor node set Q.

[0073] Among them, bidirectional traversal limits the search range by starting from the two nodes of the field pair respectively. For example, the breadth-first search algorithm is used to traverse only the one-hop neighborhood, avoiding the complexity caused by multi-hop search. Intersection operation filters nodes that are only associated with two fields at the same time through set operations, such as by comparing the intersection function built into the database or hash table.

[0074] Specifically, when a field pair (p, q) is identified as a highly correlated field pair, a set P is generated by traversing the nodes directly connected to p, and a set Q is generated by traversing the nodes directly connected to q. By calculating P∩Q, only nodes that are directly correlated with both p and q are retained. For example, the common correlated node for the geological structure field and the land area field might be the soil type field. This method excludes nodes that are only correlated with a single field, ensuring that the nodes injected into the extended field set have a strong correlation with the highly correlated field pair, thereby improving the accuracy of subsequent risk value calculations and avoiding global risk entropy bias caused by the introduction of irrelevant nodes.

[0075] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0076] The extraction process of the common associated node set includes: performing bidirectional traversal in the knowledge graph database: for field pair (p, q), obtaining the 1-hop neighbor node set P starting from p and the 1-hop neighbor node set Q starting from q; the common associated node set is the intersection of the neighbor node set P and the neighbor node set Q.

[0077] For example, in a natural resource data knowledge graph, suppose the field pair (p, q) represents "geographic coordinates" and "land area" respectively. First, starting from the "geographic coordinates" node, obtain its one-hop neighbor node set P, including "longitude," "latitude," and "altitude," etc. Then, starting from the "land area" node, obtain its one-hop neighbor node set Q, including "land use type," "area calculation method," and "plot boundary," etc. Finally, calculate the intersection of P and Q to obtain the set of commonly associated nodes, which may include nodes such as "plot boundary."

[0078] Through the above technical solution, this application achieves efficient identification and extraction of common association nodes between field pairs. By employing bidirectional traversal and set intersection operations, potential information association points are accurately located, providing crucial evidence for subsequent risk assessment and data protection. This method avoids the computational overhead of traversing the entire knowledge graph, improving the efficiency of association analysis. Simultaneously, by identifying common association nodes, it reveals potential information derivation paths between different fields, providing important support for dynamically adjusting data access strategies and implementing precise data anonymization.

[0079] In some of the solutions described above in this application, the iterative injection process of the extended field set may lead to infinite loops or redundant node injections, affecting system processing efficiency and increasing computational resource consumption. Specifically, in the process of injecting the common associated node set multiple times, if there is no effective termination mechanism, a large number of duplicate nodes may be injected repeatedly, causing the size of the extended field set to run out of control and making it impossible to complete the risk assessment within a reasonable time.

[0080] This application further proposes preset termination conditions, including the current number of iterations for the extended field set reaching a preset upper limit or the overlap between the current injected common related node set and the extended field set exceeding a preset overlap threshold.

[0081] The iteration limit sets a maximum number of iterations to restrict the growth of the extended field set, preventing system resource exhaustion due to infinite loops. The overlap threshold determines the validity of newly added nodes by calculating the intersection ratio between the common associated node set and the extended field set. For example, with a preset iteration limit of 5 and an overlap threshold of 80%, when 80% of the nodes in the common associated node set already exist in the extended field set, it is considered redundant injection and the process terminates. These two conditions form a dual constraint mechanism, preventing infinite iterations and suppressing inefficient node injection.

[0082] Specifically, during the evolution of the expanded field set, the system monitors the number of iterations and node overlap in real time. When the cumulative number of iterations reaches a preset upper limit, the injection process is immediately terminated to ensure that the processing is completed within a controllable time. Simultaneously, before each injection of a set of commonly associated nodes, the intersection ratio between that node set and the current expanded field set is calculated. If this ratio exceeds an 80% threshold, it indicates that the newly added nodes lack sufficient effective information, and continuing injection would lead to increased set redundancy. Terminating the process at this point avoids invalid calculations. For example, if an injected node set contains 100 nodes, of which 85 already exist in the expanded field set, the overlap is 85%, exceeding the threshold and triggering termination. This dual termination mechanism, combining a hard limit on the number of iterations with dynamic overlap judgment, effectively balances the completeness of risk assessment with system operating efficiency.

[0083] By employing the aforementioned technical solutions, this application avoids the waste of system resources caused by infinite iteration and improves the execution efficiency of the algorithm. Simultaneously, by setting an overlap threshold, it ensures that newly added nodes effectively supplement the expanded field set, preventing the excessive accumulation of redundant information. This flexible termination mechanism ensures both the sufficiency of risk correlation analysis and controls computational complexity, achieving a balance between security and efficiency.

[0084] In some of the solutions described above in this application, dynamic risk assessment is achieved by expanding the field set and calculating the transmissible risk value. However, the sensitivity differences of different field combinations are not considered, which makes it impossible to distinguish the contribution of highly sensitive field combinations to risk when quantifying global risk, thus affecting the accuracy of the desensitization strategy.

[0085] This application further proposes the following formula for calculating global risk entropy:

[0086]

[0087] in To preset the sensitivity interference factor, The sensitivity interference factor is assigned as follows: if the field pair contains a geographic coordinate field, the sensitivity interference factor is assigned a first-level weight; if the field pair is a combination of land area and geological structure, the sensitivity interference factor is assigned a second-level weight, where the first-level risk weight is greater than the second-level risk weight.

[0088] The global risk entropy is calculated by summing the product of the transmission risk value of each field pair and the sensitivity interference factor. The sensitivity interference factor is divided into different levels based on the type of the field pair. For example, the weight of the first level is set to 0.9, and the weight of the second level is set to 0.6. (Sensitivity Interference Factor) The assignment rules are based on the differences in the sensitive attributes of field combinations. Geographic coordinates are given higher weight due to their involvement in spatial location privacy, while the combination of land area and geological structure is given the second highest weight due to its lower risk of information leakage. The weight difference between the two is achieved through preset parameters, such as those determined during system initialization based on historical leakage cases.

[0089] Specifically, when calculating the global risk entropy, we first iterate through all field pairs in the extended field set and obtain the transmission risk value for each field pair. Subsequently, the corresponding sensitivity interference factor is matched based on the composition type of the field pair. For example, the field pair (geographic coordinates, land area) includes a geographic coordinate field, and its... The first-level weight is assigned 0.9; the field pair (land type area, geological structure) is assigned a second-level weight of 0.6. Each field pair... and The global risk entropy G is obtained by multiplying and summing the results. This calculation process incorporates the differences in sensitivity attributes of different field combinations into the global risk assessment, significantly increasing the contribution of combinations containing highly sensitive fields to the global risk. For example, in two field pairs with a transmission risk value of 0.8, the contribution of the geographic coordinate-related field pair is 0.72, while the contribution of the combination of land area and geological structure is 0.48. The risk difference between the two is accurately quantified through weight allocation, thus providing a more precise decision-making basis for subsequent desensitization operations.

[0090] Through the above technical solutions, this application achieves a quantitative assessment of the security risks of natural resource data. By introducing a sensitivity interference factor, different weights are assigned to different types of field combinations, making the risk assessment more accurate. The calculation of the global risk entropy comprehensively considers the transmission risk and sensitivity between fields, providing a scientific basis for subsequent de-identification operations and effectively improving the pertinence and effectiveness of data security management.

[0091] In some of the solutions mentioned above in this application, there is a mismatch between the calculation of global risk entropy and the objectives of the data anonymization operation during the dynamic risk assessment process. Traditional methods struggle to accurately locate the key field nodes that have the greatest impact on global risk, resulting in a lack of targeted anonymization measures and an inability to effectively balance data security and availability.

[0092] This application further proposes a desensitization strategy based on degree centrality analysis, including determining whether the global risk entropy exceeds a preset desensitization threshold and adding a noise offset to the field with the highest degree centrality in the target field set.

[0093] The field with the highest degree centrality is determined as follows:

[0094] Extract all field nodes within the target field set from the knowledge graph database, and calculate the weighted degree centrality of each field node i:

[0095]

[0096] Where j is the field node directly connected to i. For the preset connection weights, To derive the risk coefficient;

[0097] choose Add a noise offset to the field with the largest value;

[0098] The rules for applying noise offsets include adding Gaussian noise to the geographic coordinate field and scaling the values ​​of the land area field proportionally.

[0099] The weighted degree centrality calculation employs a cumulative method of multiplying and summing the connection weights of field nodes in the knowledge graph with the derived risk coefficients. Connection weights reflect the strength of associations between fields, while the derived risk coefficients characterize the probability of information leakage resulting from field pair combinations. Noise offsets are processed differently depending on the field type; geographic coordinates use Gaussian noise with a mean of zero, and land area values ​​are adjusted using a fixed scaling factor.

[0100] Specifically, all field nodes within the target field set are extracted from the knowledge graph database. For each field node, the sum of the products of its connection weights with directly connected nodes and the derived risk coefficient is calculated. The field with the largest sum of products is selected as the node with the highest degree centrality, as this node contributes the most to the global risk entropy. When adding Gaussian noise to the geographic coordinate field, the standard deviation is set to 10%-20% of the original coordinate accuracy to ensure that the coordinate offset is within an acceptable error range. When scaling the land area field proportionally, the scaling factor is controlled within the range of 0.8-1.2, maintaining the statistical characteristics of the data while destroying the precise numerical values. This differentiated desensitization process reduces the direct risk of sensitive fields being leaked and suppresses the indirect derivation risk by disrupting the data correlation of key nodes.

[0101] As a preferred embodiment, the solution of this application is implemented as follows: When the global risk entropy exceeds the preset desensitization threshold, all field nodes within the target field set need to be extracted from the knowledge graph database. For each field node, its weighted degree centrality needs to be calculated. Specifically, this involves traversing all neighboring nodes directly connected to the node, multiplying the preset connection weight on each connection edge by the derived risk coefficient, and then summing the results. The field with the largest weighted degree centrality calculation result is selected as the processing object. If the field belongs to the geographic coordinate type, a Gaussian noise generator is called to generate a noise offset based on preset mean and standard deviation parameters and superimpose it onto the original coordinate value. If the field belongs to the land area type, the value is randomly increased or decreased according to the preset scaling ratio. After noise processing is completed, the modified field data needs to be returned to the user.

[0102] Through the above technical solution, this application can dynamically identify the core field nodes with the strongest risk transmission capability and apply differentiated desensitization strategies for different data types, blocking the derivation path of sensitive information while retaining the original data characteristics of non-sensitive fields. This solution effectively solves the problem that traditional static desensitization methods cannot cope with the transmission of multi-field association risks. By combining weighted degree centrality calculation with noise processing rules, a balance between security and data availability is achieved.

[0103] In some of the solutions mentioned above in this application, the mechanism for dynamically adjusting the derived risk coefficient is not yet clear, which makes it impossible for the knowledge graph base to reflect the latest risk changes in real time, which may affect the accuracy of subsequent desensitization strategies.

[0104] This application further proposes updating the derived risk coefficient when a user's request triggers the de-identification operation:

[0105]

[0106] in The preset adjustment ratio, For global risk entropy, The risk coefficient for the derivation of the field pair (p,q) is given. This is the updated derived risk coefficient.

[0107] The process of updating the derived risk coefficient is achieved by multiplying the global risk entropy by a preset proportional coefficient μ. For example, when the global risk entropy is high, the increment of the derived risk coefficient is large, and vice versa. The value of the proportional coefficient μ can be set from 0.01 to 0.1, and the specific value is dynamically configured according to the frequency of historical risk events. During implementation, after each desensitization operation is triggered, the system automatically traverses all field pairs involved in the calculation and performs a linear superposition update on their derived risk coefficients.

[0108] Specifically, when a user request triggers the de-identification operation, the system first calculates the global risk entropy G generated by the current request. This entropy value reflects the overall risk level of the field combination. For example, when G reaches 5.0 and μ is set to 0.05, the derived risk coefficient will increase by 0.25. Through the real-time update mechanism of the knowledge graph base, subsequent access requests will directly call the updated value when calculating the propagation risk value. This dynamic adjustment method causes the derivation risk coefficient of high-risk field pairs to increase cumulatively over time. For example, after three consecutive desensitization triggers, the risk coefficient of a certain geographic coordinate and geological structure combination field pair... The value is gradually increased from an initial 0.3 to 0.45, thus obtaining a higher weight for the transmitted risk value in subsequent risk assessments. The updated derived risk coefficient will be persistently stored in the knowledge graph repository to ensure that different users share a unified risk assessment benchmark.

[0109] Through the above technical solution, this application effectively solves the technical problem that traditional static risk models cannot reflect the impact of dynamic access behavior on the association relationship of sensitive fields. By linking and updating the global risk entropy and the derived risk coefficient, the knowledge graph can record the changes in potential risk paths revealed by high-frequency desensitization operations in real time, thereby improving the accuracy of subsequent risk value calculation. This mechanism enables the system to adaptively strengthen the protection of highly associated field pairs, preventing multiple legitimate access behaviors from accumulating over time to form a sensitive information derivation channel.

[0110] Example 2:

[0111] like Figure 3 As shown, a natural resource data security management system includes:

[0112] The data acquisition module is used to acquire the target field set currently requested by the user and the knowledge graph database data; the target field set includes at least one of geographic coordinates, land area, and geological structure;

[0113] The field expansion generation module is used to traverse each field in the target field set, query the corresponding set of related nodes based on the knowledge graph library, and generate an expanded field set based on each set of related nodes.

[0114] The transmission risk calculation module is used to form field pairs from any two different fields in the extended field set, and to perform weighted calculations on each field pair to obtain the corresponding transmission risk value.

[0115] The high-association node injection module is used to determine whether the average value of the transmission risk of all fields reaches a preset evolution threshold. If so, the field pairs with transmission risk values ​​greater than the preset risk threshold are selected as high-association field pairs. The module extracts a set of common association nodes that are connected to the high-association field pairs from the knowledge graph library, injects the set of common association nodes into the extended field set, and re-determines whether the evolution threshold is reached until the preset termination condition is met.

[0116] The global risk entropy calculation module is used to calculate the global risk entropy based on the transmission risk values ​​of all field pairs in the extended field set and the preset sensitivity interference factor;

[0117] The desensitization processing module is used to perform desensitization operations on the target field set according to the global risk entropy, and update the knowledge graph database data based on the desensitization operation results.

[0118] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0119] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0120] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for managing the security of natural resource data, characterized in that, Includes the following steps: Obtain the target field set currently requested by the user and the knowledge graph database data; the target field set is at least one of geographic coordinates, land area, and geological structure. Iterate through each field in the target field set, query the set of related nodes in the knowledge graph library, and generate an extended field set; The transmission risk value of a field pair is calculated by weighting any two different fields from the extended field set. Determine whether the mean of the transmission risk value of all fields reaches the preset evolution threshold. If so, select the field pairs whose transmission risk value is greater than the preset risk threshold as high-association field pairs. Extract the common association node set that is connected to the high-association field pairs from the knowledge graph library. Inject the common association node set into the extended field set and re-determine whether the evolution threshold is reached until the preset termination condition is met. Calculate the global risk entropy based on the transmission risk values ​​of all field pairs in the extended field set and the preset sensitivity interference factor; The target field set is anonymized based on the global risk entropy, and the knowledge graph database data is updated.

2. The natural resource data security management method according to claim 1, characterized in that: The process of generating the extended field set includes: Iterate through each field in the target field set and query the knowledge graph database for a set of 1-hop neighbor nodes that have a direct relationship with that field; The initial extended field set is generated by taking the union of the target field set with the set of all 1-hop neighbor nodes obtained from the query.

3. The natural resource data security management method according to claim 2, characterized in that: The calculation process for the transmission risk value includes: For the field pair (p, q), the risk coefficient is extracted from the knowledge graph database. ; Calculate the time intensity decay factor based on real-time query logs. ; Derivation of risk coefficient and time-density decay factor Normalization is performed, and the transmission risk value is calculated: in This is a preset scaling factor.

4. The natural resource data security management method according to claim 3, characterized in that: The time density attenuation factor is calculated as follows: Retrieve the query timestamp sequence of the field pair (p, q) within the most recent T hours, calculate the standard deviation σ of the timestamps, and calculate the time density decay factor: in These are the preset adjustment parameters; The initial construction method for deriving the risk coefficient is as follows: Sensitive derivation paths are extracted from a historical database of natural resource leaks. Each path contains a sequence of field pairs that led to the information leak. For each field pair (p, q) in a path, its frequency of occurrence across all paths is calculated. Calculate the initial derived risk coefficient: in The highest frequency among all field pairs.

5. A method for managing the security of natural resource data according to claim 4, characterized in that: The extraction process of the common associated node set includes: Perform bidirectional traversal in the knowledge graph: for a field pair (p, q), obtain the set of 1-hop neighbor nodes P starting from p and the set of 1-hop neighbor nodes Q starting from q; The common set of nodes is the intersection of the neighbor set P and the neighbor set Q.

6. A method for managing the security of natural resource data according to claim 5, characterized in that: The preset termination conditions include: The current iteration count for the extended field set has reached the preset upper limit, or the overlap between the common associated node set injected this time and the extended field set exceeds the preset overlap threshold.

7. A method for managing the security of natural resource data according to claim 6, characterized in that: The formula for calculating the global risk entropy is: in To preset the sensitivity interference factor, To transmit risk values; The assignment rule for the sensitivity interference factor is as follows: If the field pair contains a geographic coordinate field, the sensitivity interference factor is the first-level weight; If the field pair is a combination of land area and geological structure, then the sensitivity interference factor is the second-level weight, where the first hazard weight is greater than the second hazard weight.

8. A method for managing the security of natural resource data according to claim 7, characterized in that: The process of performing desensitization on the target field set based on the global risk entropy includes: Determine whether the global risk entropy exceeds the preset desensitization threshold; if so, add a noise offset to the field with the highest degree centrality in the target field set. The field with the highest degree centrality is determined as follows: Extract all field nodes within the target field set from the knowledge graph database, and calculate the weighted degree centrality of each field node i: Where j is the field node directly connected to i. For the preset connection weights, To derive the risk coefficient; choose Add a noise offset to the field with the largest value; The rules for applying the noise offset include: Add Gaussian noise to the geographic coordinate field; For the land area field: scale the value proportionally.

9. A method for managing the security of natural resource data according to claim 8, characterized in that: The update process for the knowledge graph database data includes: When a user's request triggers the de-identification process, the derived risk coefficient is updated: in The preset adjustment ratio, For global risk entropy, The risk coefficient for the derivation of the field pair (p,q) is given. This is the updated derived risk coefficient.

10. A natural resource data security management system, characterized in that: Using a natural resource data security management method as described in any one of claims 1 to 9, comprising: The data acquisition module is used to acquire the target field set currently requested by the user and the knowledge graph database data; the target field set includes at least one of geographic coordinates, land area, and geological structure; The field expansion generation module is used to traverse each field in the target field set, query the corresponding set of related nodes based on the knowledge graph library, and generate an expanded field set based on each set of related nodes. The transmission risk calculation module is used to form field pairs from any two different fields in the extended field set, and to perform weighted calculations on each field pair to obtain the corresponding transmission risk value. The high-association node injection module is used to determine whether the average value of the transmission risk value of all fields reaches a preset evolution threshold. If so, the field pairs with the transmission risk value greater than the preset risk threshold are selected as high-association field pairs. The module extracts a common association node set that is connected to the high-association field pairs from the knowledge graph library, injects the common association node set into the extended field set, and re-determines whether the evolution threshold is reached until the preset termination condition is met. The global risk entropy calculation module is used to calculate the global risk entropy based on the transmission risk values ​​of all field pairs in the extended field set and the preset sensitivity interference factor; The desensitization processing module is used to perform desensitization operations on the target field set according to the global risk entropy, and update the knowledge graph database data based on the desensitization operation results.

Images