An integrated data security management method and system based on dynamic de-identification and end-to-end traceability

By adopting an integrated data security management method that combines dynamic desensitization and full-chain traceability, the problems of disconnect between data introduction and desensitization strategies, poor adaptability of static desensitization, unreliable traceability, and lack of closed-loop management have been solved, thus achieving efficient, secure, and compliant data circulation.

CN123122990APending Publication Date: 2026-11-17CHINA UNICOM XIONGAN IND INTERNET CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611510909.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-28
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies in trusted circulation scenarios suffer from problems such as a disconnect between data aggregation and desensitization strategies, static desensitization being unable to adapt to dynamic needs, insufficient end-to-end traceability capabilities, lack of integrated closed-loop security management, and untrustworthy desensitization execution environments.

Method used

An integrated data security management and control method based on dynamic desensitization and full-link traceability is adopted. By receiving heterogeneous data from multiple sources, the method performs identity authentication and encryption, automatically identifies sensitive data and generates globally unique data identifiers, combines request subject attributes and environment attributes to conduct policy evaluation, performs real-time desensitization processing, executes the process in a trusted execution environment, collects full-link traceability records and writes them to trusted storage, forming an integrated management and control closed loop.

Benefits of technology

It achieves automated integration of data import and desensitization strategies, improves the coverage of sensitive data identification and the accuracy of strategies, meets differentiated desensitization needs, enhances the credibility and compliance of data circulation, and reduces reliance on manual operations and maintenance costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123122990A_ABST
    Figure CN123122990A_ABST
Patent Text Reader

Abstract

This invention discloses an integrated data security management method and system based on dynamic de-identification and full-link traceability, belonging to the field of data security management technology. The method includes: receiving multi-source heterogeneous data and performing identity authentication and transmission encryption; obtaining the request subject attributes, environmental attributes, and usage purpose of the data usage request, performing logical operations and condition matching using policy rules to obtain policy decision results and perform de-identification operations; collecting data information at key nodes and generating traceability records, which are then written to trusted storage after hashing; performing risk scoring based on the traceability records, and dynamically adjusting policy rules according to the risk identification results. This invention achieves automated integration of data aggregation and dynamic de-identification strategies, avoiding the lag of manual configuration; simultaneously, through full-link traceability and trusted storage, it ensures the traceability of data flow and the immutability of audit evidence, improving the security of data circulation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of data security management technology, specifically to an integrated data security management method and system based on dynamic desensitization and full-link traceability. Background Technology

[0002] With the development of the digital economy, the demand for data as a production factor to circulate and share among different organizations and systems is growing. In trusted circulation scenarios such as government data sharing, financial data exchange, medical data collaboration, and enterprise data transactions, data needs to be aggregated, processed, de-identified, and transferred under controlled conditions, while meeting security compliance and availability requirements. Currently, the field of data security management has developed various technical means such as data classification and grading, data de-identification, access control, and data traceability. However, existing technologies still have the following obvious defects and shortcomings in practical applications: 1. The data aggregation and import process is disconnected from the generation of de-identification strategies. In existing technologies, data import is typically handled by data access gateways or ETL tools, primarily responsible for data extraction, format conversion, and loading into the database. However, the configuration of de-identification strategies is often manually set by security administrators based on experience, separate from the data import process. After data import, sensitive fields need to be manually identified, sensitivity levels determined, and de-identification rules configured one by one. This approach has significant problems: firstly, manual identification is inefficient and prone to omissions, making it difficult to handle the rapid import of massive fields from multi-source heterogeneous data; secondly, the update of de-identification strategies lags behind changes in data content. When data sources add sensitive fields or data classification and grading standards are adjusted, the de-identification strategies cannot be automatically updated, resulting in inaccurate de-identification granularity and situations where de-identification is not performed where necessary or is over-identified. Furthermore, the lack of automated classification and grading labeling and asset registration mechanisms in the data import process prevents subsequent de-identification strategies from dynamically matching based on unified data identifiers and classification and grading labels.

[0003] 2. Static data masking methods cannot adapt to the differentiated needs of dynamic data circulation scenarios. Traditional data masking mainly adopts static masking methods, which involves a one-time, irreversible transformation of data before it is entered into the database or exported. However, in trusted circulation scenarios, the same data often needs to be circulated multiple times to different subjects, environments, and purposes. Different users have different requirements for the granularity of data visibility. For example, internal employees may be allowed to view the complete ID number, while external partners may only be allowed to view the masked portion of the information; trusted terminals on the internal network may allow the return of plaintext, while untrusted terminals on the external network must be forcibly masked. Static masking processes data uniformly and cannot be differentiated and dynamically adjusted according to factors such as the requesting subject's attributes, environmental attributes, and usage purpose. This can easily lead to an over-masking phenomenon of "one-size-fits-all," resulting in decreased data availability and even leakage of sensitive information due to insufficient data masking. Although some dynamic de-identification solutions exist in existing technologies, their strategy evaluation dimensions are relatively simple, mainly based on user roles or data table levels for control. They lack deep integration with multi-dimensional attributes such as data classification and grading labels, environmental context, and usage purpose, and cannot finely match de-identification algorithms and de-identification granularity after strategy decision-making.

[0004] 3. Insufficient data traceability capabilities hinder end-to-end tracking and compliance auditing in trusted data circulation scenarios. Existing data traceability technologies are mostly limited to operation logs within a single system, such as database audit logs and application access logs. These logs often suffer from the following problems: First, logs are scattered across different systems, lacking a unified data identifier for association, making it impossible to establish a complete data lineage across business and systems, and difficult to answer key questions such as "where did the data come from, where did it go, and where did it flow through?" Second, log recording granularity is coarse, typically only recording operation type, time, and user, lacking records of key information such as the version of the de-identification strategy, de-identification algorithm, operation context, and environmental attributes, making it impossible to accurately reconstruct the data processing state at the time of the incident. Third, log records are easily tampered with or deleted, lacking a reliable storage mechanism, and are difficult to use as reliable evidence for compliance auditing and dispute resolution. In trusted data circulation scenarios, data often involves multiple parties. Once data leakage or unauthorized use occurs, it is necessary to trace back to the specific link and responsible person, and existing data traceability capabilities are clearly insufficient.

[0005] 4. Data security management is fragmented across different stages, lacking an integrated closed-loop mechanism. In existing technologies, data aggregation and import, data anonymization, data traceability, and audit control are typically deployed as independent functional modules, completed separately by different systems or tools. The classification and grading results generated during data import cannot be automatically passed to the anonymization engine; records generated during the anonymization process cannot be seamlessly integrated with the traceability system; and abnormal behaviors and risk events discovered through audit analysis cannot provide real-time feedback to adjust the anonymization strategy. This fragmentation leads to a significant lag in security policy response: for example, when traceability analysis reveals that a certain type of data is being exported abnormally at high frequencies, the system cannot automatically increase the anonymization intensity of that data or trigger an approval process; manual intervention is still required to adjust the strategy. Furthermore, the anonymization strategy itself lacks version management, and the history of strategy changes cannot be traced, making it difficult to meet compliance audit requirements. Overall, existing technologies cannot form a closed loop of "data aggregation and import → dynamic anonymization → full-link traceability → integrated management," making it difficult to address the dynamically changing security threats and compliance requirements of data circulation.

[0006] 5. The de-identification process lacks the protection of a trusted execution environment, posing a risk of plaintext data leakage. In existing dynamic de-identification solutions, de-identification operations are typically performed on ordinary application servers or database proxies. The policy decision-making process and plaintext data may be exposed to an untrusted operating environment, making them vulnerable to memory attacks, unauthorized internal operations, or system vulnerabilities, leading to the leakage of plaintext sensitive data or the bypassing of de-identification strategies. Although some solutions employ encryption and access control measures, they do not fundamentally isolate the de-identification operations from the plaintext data, failing to guarantee the trustworthiness of the de-identification process itself.

[0007] In summary, existing technologies for data security management in trusted data circulation scenarios have several prominent shortcomings, including a disconnect between data aggregation and desensitization strategies, static desensitization's inability to adapt to dynamic needs, insufficient end-to-end traceability capabilities, a lack of integrated closed-loop security management, and an untrustworthy desensitization execution environment. Summary of the Invention

[0008] This invention addresses the shortcomings of existing technologies by proposing an integrated method and system for dynamic desensitization and end-to-end traceability of data introduction in a trusted circulation scenario. It aims to solve problems such as the disconnect between data introduction and desensitization, poor desensitization adaptability, unreliable traceability, lack of closed-loop control, and insecure desensitization execution environment in existing technologies.

[0009] The technical solution provided by this invention is as follows: An integrated data security management method based on dynamic de-identification and full-chain traceability includes: Step 1: Receive heterogeneous data from multiple sources from the data source layer, authenticate the data source, encrypt the transmission channel, perform protocol adaptation and format conversion to obtain standardized data; identify sensitive data, classify and label the standardized data, generate globally unique data identifiers, and complete data asset registration. Step 2: When a data user initiates a data usage request, the request subject attribute, environment attribute, and usage purpose corresponding to the data usage request are obtained, and together with the classification and grading label corresponding to the globally unique data identifier, they are used as policy evaluation attributes. Then, logical operations and condition matching are performed according to the preset policy rules to obtain the policy decision result. The policy decision result is sent to the trusted execution environment to perform real-time de-identification processing on the data to be de-identified. Step 3: Collect operational behavior information and data flow information at key nodes in the data aggregation and introduction, dynamic desensitization and data circulation process, and standardize the collected information into traceability records; perform hash operation on the traceability records and write them into trusted storage, and use the globally unique data identifier to associate multiple traceability records to form a data lineage diagram; Step 4: The integrated management and control module reads the traceability records from the trusted storage, performs security audits and risk identification on the data flow process, then dynamically adjusts the preset policy rules based on the risk identification results, and writes the policy change records into the trusted storage to form an integrated management and control closed loop.

[0010] Furthermore, in step 1, the standardized data is subjected to sensitive data identification, classification, and hierarchical labeling, including: The system employs at least one of rule matching, natural language processing, and machine learning to identify sensitive data in fields, records, and files. Based on preset classification and grading standards, the identified sensitive data is categorized and graded, generating classification and grading labels.

[0011] Furthermore, in step 2, the request subject attributes include user identity, role, department, permission level, and authentication strength; the environmental attributes include network location, terminal security status, time, location, and device identifier; and the purpose of use includes statistical analysis, model training, business processing, and compliance auditing.

[0012] Furthermore, in step 2, the strategy decision result is sent to the trusted execution environment to perform real-time de-identification processing on the data to be de-identified, including: selecting at least one de-identification algorithm from the algorithm library based on the strategy decision result, such as masking, randomization, generalization, differential privacy, and format-preserving encryption, and determining the range of de-identified fields and the de-identification granularity, and then performing data de-identification processing in the trusted execution environment.

[0013] Furthermore, in step 3, the collected operation behavior information and data flow information include: operation subject, operation type, operation time, data identifier, data source, data destination, de-identification strategy version, de-identification algorithm, de-identification fields, environmental attributes, and request purpose.

[0014] Furthermore, in step 3, the data structure of the generated traceability record includes: record identifier, data identifier, operation subject, operation type, timestamp, desensitization strategy version, upstream data identifier, downstream data identifier, environmental attributes, purpose of use, hash value, and extended fields; wherein, the upstream data identifier and downstream data identifier are used to associate the upstream and downstream relationships of data flow.

[0015] Furthermore, in step 3, the traceability record is hashed and then written into trusted storage. Specifically, the trusted storage adopts a blockchain network or a trusted storage system based on trusted timestamp services. The blockchain network is used to package the hash value of the traceability record onto the chain to ensure that the record is immutable and verifiable. The trusted storage system is used to add a trusted timestamp to each record.

[0016] Furthermore, in step 4, a security audit and risk identification are performed on the data flow process, and then the preset policy rules are dynamically adjusted based on the risk identification results. Specifically, this includes: Detect risk events during data flow, including abnormal access behavior, unauthorized operations, high-frequency export, and policy bypass; score the detected risk events; and trigger dynamic control measures when the risk score exceeds a preset threshold, dynamically adjusting the preset policy rules.

[0017] Furthermore, in step 4, the preset strategy rules are dynamically adjusted, including at least one of the following: dynamically adjusting the desensitization strategy parameters, restricting the circulation of datasets, adjusting the rules for sending alarm information, or adding an approval process.

[0018] This invention also provides an integrated data security management system based on dynamic desensitization and full-link traceability, used to execute the above-mentioned integrated data security management method, including: The data source layer contains various data sources to provide the raw data to be circulated. The trusted data access layer, including the trusted data access gateway, is used for unified access, protocol adaptation, identity authentication, transmission encryption, and format conversion of the data source layer; The data processing and desensitization layer includes a classification, grading, labeling, and metadata management module; a dynamic desensitization strategy engine; and a trusted execution environment. The classification, grading, labeling, and metadata management module identifies sensitive data in standardized data, classifies and labels it, generates globally unique data identifiers, and registers data assets. It then provides the classification and grading labels and data identifiers to the dynamic desensitization strategy engine. The dynamic desensitization strategy engine evaluates and matches strategies based on the subject attributes, environment attributes, purpose of use, and classification and grading labels of the data usage request, determining the desensitization algorithm, the scope of desensitized fields, and the desensitization granularity. The trusted execution environment performs real-time desensitization processing on the data to be desensitized within the trusted execution environment, based on a hardware root of trust. The traceability and control layer includes a full-link traceability collection module, a trusted storage module, and an integrated control module. The full-link traceability collection module is used to collect operational behavior information and data flow information at key nodes during data introduction, dynamic desensitization, and data circulation, generating traceability records. The trusted storage module is used to receive traceability records after hashing and perform trusted storage. The integrated control module is used to read traceability records from the trusted storage module for security auditing and risk identification, and dynamically adjusts preset policy rules when the risk score exceeds a preset threshold, feeding back the policy adjustment results to the dynamic desensitization policy engine to form an integrated control closed loop.

[0019] Compared with the prior art, the present invention discloses the following technical effects: This invention, upon completion of data access at the trusted data access gateway, immediately utilizes rule matching, natural language processing, and machine learning to automatically identify sensitive data. It classifies and labels data according to preset standards, generating globally unique data identifiers that persist throughout the entire data lifecycle. Since the classification and labeling tags and data identifiers are directly provided to the dynamic data masking strategy engine, the masking strategy does not require manual configuration of each item; instead, it is based on automatic tag matching, avoiding the subjectivity and lag of manual identification and significantly reducing labor costs. It also improves the coverage of sensitive data identification and the accuracy of the masking strategy. When the data source content or classification and grading standards change, the system can update the tags promptly during the data import stage, and the strategy engine automatically adjusts synchronously, achieving real-time strategy updates and solving the problem of strategy configuration lagging behind data changes in existing technologies.

[0020] Existing static data masking methods, after uniform data processing, cannot adapt to the differentiated needs of different subjects, environments, and purposes, easily leading to over- or under-masking. This invention uses request subject attributes, environmental attributes, usage purpose, and data classification and grading tags as policy evaluation attributes. The policy engine performs logical operations and condition matching according to preset rules, dynamically selecting masking algorithms such as masking, randomization, generalization, differential privacy, and format-preserving encryption from the algorithm library, and determining the scope and granularity of the masked fields. This allows the same data to automatically adopt different masking strategies when facing internal employees versus external partners, trusted intranet terminals versus untrusted extranet terminals, and statistical analysis versus model training purposes. This maximizes data usability while ensuring sensitive information is not leaked, avoiding the data value loss caused by a "one-size-fits-all" masking approach. Simultaneously, real-time masking processing is executed in a hardware-based trusted execution environment, where policy decisions and plaintext data are isolated within trusted boundaries, effectively preventing memory attacks, unauthorized internal operations, and policy bypassing, further enhancing the security of the masking process.

[0021] This invention utilizes a full-link traceability acquisition module to uniformly embed data at key nodes during data introduction, dynamic desensitization, and data circulation. This collects fine-grained information such as the operating entity, operation type, timestamp, desensitization strategy version, upstream data identifier, downstream data identifier, environmental attributes, and purpose of use, generating standardized traceability records. These records are then hashed and written to a blockchain network or a trusted storage system based on trusted timestamps, ensuring the immutability and verifiability of the records. Globally unique data identifiers are used to link each record, forming a complete data lineage diagram, enabling traceability and auditing of data flow at any stage. This improvement addresses the shortcomings of existing technologies, such as scattered logs, susceptibility to tampering, and inability to correlate data across stages. It provides reliable evidence for data leakage liability determination, compliance auditing, and dispute resolution, significantly enhancing the credibility of data circulation.

[0022] In existing technologies, data import, desensitization, tracing, and auditing modules are fragmented, and abnormal behaviors detected by audits cannot automatically trigger policy adjustments, resulting in a significant lag in security response. The integrated management module of this invention reads tracing records from trusted storage, detects risky behaviors such as abnormal access, unauthorized operations, high-frequency exports, and policy bypassing based on behavioral baselines, and performs risk scoring. When the risk score exceeds a preset threshold, the system automatically triggers control measures such as dynamic adjustment of desensitization policy parameters, data blocking, secondary approval, or alarms, and writes the policy change record to trusted storage. This closed-loop feedback mechanism allows security policies to be adjusted in real time according to actual circulation risks without manual intervention, significantly shortening the response time from risk detection to policy implementation. Simultaneously, policy version management and retrospective functions ensure the auditability and reversibility of the dynamic policy adjustment process. Compared to the isolated operation of each link in existing technologies, this invention achieves a shift in security management from passive response to proactive defense, significantly improving the overall security protection level and operational efficiency.

[0023] The various functional modules of this invention are connected via standard interfaces. The data source layer, trusted data access layer, dynamic de-identification strategy engine, trusted execution environment, end-to-end traceability and collection module, trusted storage module, and integrated management and control module can be deployed independently and flexibly combined without requiring large-scale modifications to the original data source system. The entire process of data aggregation and import, dynamic de-identification, end-to-end traceability, and management is automated, reducing reliance on manual operations and lowering operational costs and management complexity. Furthermore, the de-identification algorithm library and policy rules support flexible expansion and can be quickly adapted to different industries and compliance requirements, exhibiting good versatility and scalability.

[0024] In summary, this invention, through a series of technological improvements such as automated classification and grading labeling, dynamic desensitization of multi-dimensional attributes, trusted execution environment protection, end-to-end trusted traceability, and integrated closed-loop management, effectively overcomes the shortcomings of existing technologies, such as the disconnect between data introduction and desensitization, poor adaptability of static desensitization, insufficient traceability capabilities, lack of closed-loop management, and untrustworthy execution environment. While improving the security and compliance of data circulation, it reduces implementation and maintenance costs, and improves data availability and ease of use, demonstrating significant practical value and promising prospects for widespread application. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments 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.

[0026] Figure 1 This is a diagram of the integrated data security management system architecture provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the integrated data security management and control method provided in an embodiment of the present invention; Figure 3 This is a diagram of the end-to-end traceability record data structure provided in an embodiment of the present invention; Figure 4 This is a system deployment architecture diagram for a government data sharing scenario provided in an embodiment of the present invention; Figure 5 This is a flowchart illustrating the data flow and anonymization process in a government data sharing scenario provided in this embodiment of the invention. Figure 6 This is a schematic diagram illustrating the traceability and closed-loop management of government data sharing scenarios provided in this embodiment of the invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, 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] This invention provides an integrated data security management system based on dynamic de-identification and end-to-end traceability, such as... Figure 1 As shown, it includes the following functional modules: The data source layer includes various data sources such as relational databases, non-relational databases, file systems, message queues, and data interfaces, which are used to provide the raw data to be circulated.

[0029] The trusted data access layer includes a trusted data access gateway, which is used for unified access, protocol adaptation, identity authentication, transmission encryption, and format conversion of the data source layer. The trusted data access gateway connects to the data source layer through a standard data interface to authenticate the identity of the data source and encrypt the transmission channel, ensuring the confidentiality and integrity of the data aggregation and introduction process.

[0030] The data processing and desensitization layer includes a classification and grading labeling and metadata management module, a dynamic desensitization strategy engine, and a trusted execution environment; The classification, grading, labeling, and metadata management module connects to the trusted data access gateway, receives standardized data, identifies sensitive data, classifies and labels the data, generates globally unique data identifiers, and completes data asset registration. This module provides metadata information such as classification and grading labels and data identifiers to the dynamic desensitization strategy engine. The dynamic desensitization strategy engine connects with the classification and hierarchical labeling and metadata management module. It receives the subject attributes, environment attributes and usage purpose in the data usage request, combines them with the classification and hierarchical labels in the metadata, and performs strategy evaluation and matching according to preset strategy rules to determine the desensitization algorithm, the scope of desensitized fields and the desensitization granularity. The dynamic desensitization strategy engine then sends the desensitization strategy decision results to the trusted execution environment. The Trusted Execution Environment (TEE) connects with the dynamic de-identification strategy engine, receives the de-identification strategy decision results, performs real-time de-identification processing on the data to be de-identified within the TEE, and embeds a traceable watermark. The TEE is built on a hardware root of trust, isolating the strategy decision process, the de-identification operation process, and the plaintext data to prevent unauthorized access and policy bypass.

[0031] The traceability and control layer includes a full-chain traceability and data collection module, a trusted storage module, and an integrated control module; The end-to-end traceability and data collection module is used to collect operational behavior information and data flow information at key nodes in the data import, dynamic desensitization and data circulation process, and generate traceability records. This module is connected to the trusted data access gateway, the classification and hierarchical labeling and metadata management module, the dynamic desensitization strategy engine, the trusted execution environment and the data consumption side, and collects data at each key node. The trusted storage module connects to the end-to-end traceability acquisition module, receiving traceability records and writing them into the trusted storage. The trusted storage can be a blockchain network or a trusted storage system based on trusted timestamp services. Traceability records are written after hash calculations to ensure immutability and verifiability. The integrated management and control module connects to the trusted storage module to read traceability records from the trusted storage module and perform security audits and risk identification. Then, based on the risk identification results, it dynamically adjusts the preset policy rules in the de-identification policy engine and feeds back the policy adjustment results to the dynamic de-identification policy engine, forming an integrated management and control closed loop. At the same time, it also has policy version management and rollback functions, which support viewing the policy status at any point in time during compliance audits and quickly rolling back when policy adjustments cause anomalies.

[0032] This invention also provides an integrated data security management method based on dynamic desensitization and full-chain traceability, applicable to the aforementioned integrated data security management system, such as... Figure 2 As shown, it includes: Step 1: The trusted data access gateway receives multi-source heterogeneous data from the data source layer. The access gateway supports various data source protocols, including JDBC / ODBC, RESTful API, file transfer, and message subscription. During the access process, the data source identity is authenticated, and the transmission channel is encrypted to ensure the confidentiality and integrity of the imported data. After being accessed, the data enters the classification, grading, labeling, and metadata management module for sensitive data identification and classification / grading labeling. Specifically, this invention employs at least one method among rule matching, natural language processing, and machine learning to identify sensitive data in fields, records, and files. For example, it uses regular expressions to identify ID card numbers, mobile phone numbers, and bank card numbers; named entities to identify names, addresses, and organization names; and classification models to identify sensitive semantic content in text. Based on preset classification and grading standards, the identified sensitive data is categorized and graded, generating classification and grading labels.

[0033] A globally unique data identifier is generated for each dataset. This identifier is used throughout the entire data lifecycle—from data aggregation and introduction to anonymization, circulation, and use—to achieve end-to-end data traceability. Simultaneously, data quality is assessed, including completeness, accuracy, consistency, and timeliness, and the assessment results are written into the metadata. The metadata also includes information such as data source, data format, data owner, and update time. After data asset registration is completed, classification and grading labels and data identifiers are provided to the dynamic anonymization strategy engine.

[0034] Step 2: When a data user initiates a data query, data export, or data sharing request, the system obtains the request subject attributes, environment attributes, and usage purpose corresponding to the data usage request. Request subject attributes include user identity, role, department, permission level, authentication strength, etc.; environment attributes include network location, terminal security status, time, location, device identifier, etc.; usage purpose includes statistical analysis, model training, business processing, compliance auditing, etc.

[0035] The dynamic data masking strategy engine is built upon an attribute access control model. The engine uses the request subject attributes, environment attributes, purpose of use, and data identifier corresponding to classification and hierarchical labels as policy evaluation attributes. Based on preset policy rules, it performs logical operations and condition matching to obtain the policy decision result. For example, the policy rule can be expressed as follows: if the data level is "confidential," the request subject role is "external partner," and the purpose of use is "model training," then differential privacy masking is performed on the ID number field, mask masking is performed on the name field, and generalized masking is performed on the address field; if the data level is "internal," the request subject role is "internal employee," and the environment attribute is "trusted intranet terminal," then plaintext data is allowed to be returned within a limited time, but a watermark must be added. Based on the policy decision result, at least one masking algorithm from the algorithm library is selected, including masking, randomization, generalization, differential privacy, and format-preserving encryption, and the range and granularity of the masked fields are determined. For example, the mask retains the first 3 and last 4 characters, and the middle character is replaced with "". Differential privacy satisfies ε-differential privacy, and the privacy budget ε is dynamically set according to the data sensitivity level.

[0036] Real-time data masking is performed in a trusted execution environment, which can be SGX, TrustZone, SEV, etc. The policy decision-making process and the data masking operation are both completed within the trusted execution environment, and the plaintext data does not leave the trusted boundary to prevent unauthorized access and policy bypass.

[0037] The anonymized data can be embedded with a traceable watermark. This watermark is associated with a globally unique data identifier and the requesting entity corresponding to the current data usage request, enabling the identification and tracing of the source of a data breach. After the anonymization process is complete, the anonymized data is output to the data consumer.

[0038] Step 3: During the data aggregation and import, dynamic desensitization, and data circulation processes, the end-to-end traceability and data collection module collects operational behavior information and data flow information at key nodes. The collected content includes: operation subject, operation type, operation time, data identifier, data source, data destination, desensitization strategy version, desensitization algorithm, desensitized fields, environmental attributes, and request purpose.

[0039] The source tracing and data collection module standardizes the collected information into source tracing records (see reference). Figure 3 The data structure for tracing records includes: record identifier, data identifier, operation subject, operation type, timestamp, anonymization strategy version, upstream data identifier, downstream data identifier, environmental attributes, purpose of use, hash value, and extended fields. Among these, the upstream and downstream data identifiers are used to associate the upstream and downstream relationships in the data flow.

[0040] The traceability records are hashed and then written to trusted storage. Trusted storage can utilize a blockchain network to package the hash values ​​of the traceability records onto the chain, ensuring the records are immutable and verifiable; alternatively, it can employ a trusted storage system based on trusted timestamp services, adding a trusted timestamp to each record. Then, globally unique data identifiers are used to link multiple traceability records, forming a chain relationship and constructing a data lineage graph. This graph displays the complete path and transformation relationships of data from its introduction, de-identification, circulation, to its use.

[0041] Step 4: The integrated management and control module reads standardized traceability records from trusted storage to perform security audits and risk identification on the data flow process. Security audits include access behavior audits, policy execution audits, data export audits, and data flow compliance audits. Risk identification specifically includes: building a behavioral baseline based on traceability records to detect abnormal access behavior, unauthorized operations, high-frequency exports, policy bypassing, and other risky behaviors. Detected risk events are scored, taking into account factors such as behavioral deviation, data sensitivity level, scope of impact, and historical behavioral patterns.

[0042] When the risk score exceeds a preset threshold, dynamic control measures are triggered, including at least one of the following: dynamically adjusting the desensitization strategy parameters, restricting the flow of datasets, adjusting the rules for sending alarm information, or adding an approval process. For example, adjusting desensitization strategy parameters could involve increasing the desensitization intensity, adding an approval process, or limiting the amount of data exported; data blocking could involve suspending the flow of relevant entities or datasets; alarm notification could involve sending alarm information to the data security administrator; and secondary approval could require additional manual approval for the flow of highly sensitive data.

[0043] After the strategy is adjusted, the dynamic desensitization strategy engine generates a new strategy version and writes the strategy change record into trusted storage, forming an integrated management and control closed loop.

[0044] In summary, this invention achieves automated integration of data aggregation and dynamic desensitization strategies, avoiding the lag and inaccuracy of manual configuration; it enables dynamic desensitization based on multi-dimensional attributes, meeting the differentiated security needs of different entities, environments, and purposes; through end-to-end traceability and trusted storage, it ensures the traceability of data flow and the immutability of audit evidence; and around a closed-loop feedback mechanism for control, it enables real-time dynamic adjustment of security strategies, significantly improving the security and compliance of data circulation.

[0045] This invention also uses a municipal government data sharing platform as an application scenario to further illustrate the role of the above-mentioned integrated data security management and control method in the trusted circulation of government data.

[0046] 1. System deployment.

[0047] The government data sharing platform deploys a trusted data access gateway, a classification and hierarchical labeling and metadata management module, a dynamic de-identification strategy engine, a trusted execution environment, a full-link traceability and data collection module, a trusted storage module, and an integrated management and control module. The trusted execution environment is built on Intel SGX; the trusted storage module uses a consortium blockchain network, with nodes jointly maintained by participating government departments; the data source layer includes the business databases, file servers, and message queues of various commissions and bureaus. The system deployment architecture is as follows: Figure 4 As shown.

[0048] 2. Data introduction.

[0049] The trusted data access gateway connects to heterogeneous data sources, including the marriage registration database of the Civil Affairs Bureau, the population information database of the Public Security Bureau, and the health record database of the Health Commission, via a JDBC interface. During data access, the data source identity undergoes two-way TLS authentication, and the transmission channel is encrypted using the national cryptographic algorithm SM4. After data access, it enters the classification, grading, labeling, and metadata management module. This module uses a regular expression rule base to identify sensitive fields such as ID card numbers, mobile phone numbers, and addresses; it uses the BERT named entity recognition model to identify entities such as names of persons and organizations; and it uses the FastText classification model to perform sensitive semantic classification on text annotation fields. According to the "Guidelines for the Grading and Classification of Government Data," the data is divided into four levels: L1 Public, L2 Internal, L3 Sensitive, and L4 Confidential, and labeled with data categories such as "Personal Identity Information" and "Health Information." For example, in marriage registration data, the "Name" field is labeled as L3 Personal Identity Information, the "ID Card Number" as L4 Personal Identity Information, and the "Registration Date" as L1 Public. The system generates a globally unique data identifier for each dataset, such as "ZW-20260901-001," and writes it to the metadata management database to complete the data asset registration. Data quality assessment results show a completeness of 98.5% and an accuracy of 99.2%, meeting circulation requirements.

[0050] 3. Dynamic desensitization.

[0051] When a staff member of a district administrative approval bureau initiates a data query request through the data sharing platform to obtain marital status information of retired personnel within the jurisdiction for business processing, the system obtains the requester's attributes (role: district-level approval personnel; department: administrative approval bureau; permission level: L3 authorization), environmental attributes (network location: government extranet; terminal security status: compliant terminal management software installed; time: weekdays 8:30-18:00), and purpose of use (required for business processing). The dynamic de-identification strategy engine jointly evaluates the above attributes with the classification and grading labels corresponding to the data identifier ZW-20260901-001. The preset strategy rules stipulate that when the data level is L3 or L4, the purpose of use is business processing, and the environment is a compliant terminal on the government extranet, the L4 level field (ID number) is subjected to format preservation encryption de-identification, requiring the retention of the first 6 digits and the last 4 digits, with the middle filled with ciphertext; the L3 level field (name) is subjected to mask de-identification, retaining the surname, and the given name is masked. "Instead, the strategy decision results are sent to the trusted execution environment. Real-time de-identification is performed on the query result set within the SGX isolation environment, ensuring that plaintext data does not leave the trusted execution environment. After data de-identification is complete, the system embeds an invisible watermark in the result data. The watermark information includes the data identifier ZW-20260901-001 and the request subject ID, used for leak tracing and source tracking. The data flow and de-identification process in the government data sharing scenario is as follows..." Figure 5 As shown.

[0052] 4. End-to-end traceability.

[0053] The end-to-end traceability and data collection module collects information at each stage, including data access, classification and grading, strategy matching, data masking, and data output. The traceability record structure is as follows: Figure 6 As shown.

[0054] For example, a traceability record might be as follows: Record ID: SY-20260901-001; Data ID: ZW-20260901-001; Operating Entity: “District Administrative Approval Bureau - Zhang San”; Operation Type: “Data Query and Anonymized Output”; Timestamp: “2026-09-01 10:24:56.121”; Anonymization Strategy Version: “POL-V3.2”; Upstream Data ID: “ZW-20260901-001”; Downstream Data ID: “ZW-20260901-001-Copy”; Environmental Attribute: “Government Affairs Extranet - Compliant Terminal”; Purpose of Use: “Business Processing Required”; Hash value calculated using SHA-256. This record, after being hashed, is written to the consortium blockchain and stored synchronously by all participating nodes.

[0055] 5. Integrated management and control.

[0056] The integrated management and control module periodically reads traceability records from the blockchain for audit analysis. By constructing historical behavior baselines for each entity, it was detected that an account frequently exported L3-level data during non-working hours, reaching a risk score of 65, exceeding the preset threshold of 60. The system automatically triggered control measures: temporarily downgrading the account's query permissions, forcibly increasing the anonymization strength for subsequent requests (only the first 3 digits of the ID number are retained), and sending an alert to the data security administrator. After the policy adjustment, the policy engine generates a new version, POL-V3.3, and the change record is written to the blockchain, achieving policy version traceability.

[0057] In this embodiment, the automatic identification accuracy rate of the data introduction process reaches 97.6%, the average time for desensitization strategy matching is 15 milliseconds, and the latency of uploading traceability records to the blockchain is less than 1 second. The overall process realizes a closed loop of "data aggregation and introduction → dynamic desensitization → full-link traceability → integrated management and control", which significantly improves the security management level of government data sharing.

[0058] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A data security management method based on dynamic desensitization and full-chain traceability, characterized in that, Includes the following steps: Step 1: Receive heterogeneous data from multiple sources from the data source layer, authenticate the data source, encrypt the transmission channel, perform protocol adaptation and format conversion to obtain standardized data; identify sensitive data, classify and label the standardized data, generate globally unique data identifiers, and complete data asset registration. Step 2: When a data user initiates a data usage request, the request subject attribute, environment attribute, and usage purpose corresponding to the data usage request are obtained, and together with the classification and grading label corresponding to the globally unique data identifier, they are used as policy evaluation attributes. Then, logical operations and condition matching are performed according to the preset policy rules to obtain the policy decision result. The policy decision result is sent to the trusted execution environment to perform real-time de-identification processing on the data to be de-identified. Step 3: Collect operational behavior information and data flow information at key nodes in the data aggregation and introduction, dynamic desensitization and data circulation process, and standardize the collected information into traceability records; perform hash operation on the standardized traceability records and write them into trusted storage, and use the globally unique data identifier to associate multiple traceability records to form a data lineage diagram; Step 4: The integrated management and control module reads the traceability records from the trusted storage, performs security audits and risk identification on the data flow process, then dynamically adjusts the preset policy rules based on the risk identification results, and writes the policy change records into the trusted storage to form an integrated management and control closed loop.

2. The integrated data security management and control method according to claim 1, characterized in that, Step 1 involves identifying, classifying, and labeling standardized data as sensitive data, including: The system employs at least one of rule matching, natural language processing, and machine learning to identify sensitive data in fields, records, and files. Based on preset classification and grading standards, the identified sensitive data is categorized and graded, generating classification and grading labels.

3. The integrated data security management and control method according to claim 1, characterized in that, In step 2, the request subject attributes include user identity, role, department, permission level, and authentication strength; the environmental attributes include network location, terminal security status, time, location, and device identifier; and the purpose of use includes statistical analysis, model training, business processing, and compliance auditing.

4. The integrated data security management and control method according to claim 3, characterized in that, In step 2, the strategy decision result is sent to the trusted execution environment to perform real-time de-identification processing on the data to be de-identified. This includes: selecting at least one de-identification algorithm from the algorithm library based on the strategy decision result, such as masking, randomization, generalization, differential privacy, and format-preserving encryption, and determining the range and granularity of the de-identified fields. Then, the data de-identification processing is performed in the trusted execution environment.

5. The integrated data security management and control method according to claim 1, characterized in that, In step 3, the collected operation behavior information and data flow information include: operation subject, operation type, operation time, data identifier, data source, data destination, de-identification strategy version, de-identification algorithm, de-identification fields, environmental attributes, and request purpose.

6. The integrated data security management and control method according to claim 5, characterized in that, In step 3, the data structure of the traceability record includes: record identifier, data identifier, operation subject, operation type, timestamp, desensitization strategy version, upstream data identifier, downstream data identifier, environmental attributes, purpose of use, hash value, and extended fields; among them, the upstream data identifier and downstream data identifier are used to associate the upstream and downstream relationships of data flow.

7. The integrated data security management and control method according to claim 6, characterized in that, In step 3, the traceability record is hashed and then written into trusted storage. Specifically, the trusted storage adopts a blockchain network or a trusted storage system based on trusted timestamp services. The blockchain network is used to package the hash value of the traceability record onto the chain to ensure that the record is immutable and verifiable. The trusted storage system is used to add a trusted timestamp to each record.

8. The integrated data security management and control method according to claim 1, characterized in that, In step 4, a security audit and risk identification are performed on the data flow process, and then the preset policy rules are dynamically adjusted based on the risk identification results. Specifically, this includes: Detect risk events during data flow, including abnormal access behavior, unauthorized operations, high-frequency export, and policy bypass; score the detected risk events; and trigger dynamic control measures when the risk score exceeds a preset threshold, dynamically adjusting the preset policy rules.

9. The integrated data security management and control method according to claim 8, characterized in that, In step 4, the preset strategy rules are dynamically adjusted, including at least one of the following: dynamically adjusting the desensitization strategy parameters, restricting the circulation of datasets, adjusting the rules for sending alarm information, or adding an approval process.

10. An integrated data security management and control system based on dynamic desensitization and full-link traceability, used to execute the integrated data security management and control method according to any one of claims 1-9, characterized in that, include: The data source layer contains various data sources to provide the raw data to be circulated. The trusted data access layer, including the trusted data access gateway, is used for unified access, protocol adaptation, identity authentication, transmission encryption, and format conversion of the data source layer; The data processing and desensitization layer includes a classification and hierarchical labeling and metadata management module, a dynamic desensitization strategy engine, and a trusted execution environment. The classification and hierarchical labeling and metadata management module is used to identify sensitive data in standardized data, classify and label it, generate globally unique data identifiers, and complete data asset registration. It also provides the classification and hierarchical labels and data identifiers to the dynamic desensitization strategy engine. The dynamic de-identification strategy engine is used to evaluate and match strategies based on the subject attributes, environment attributes, purpose of use, and classification and grading tags of the data usage request, and to determine the de-identification algorithm, the scope of de-identified fields, and the de-identification granularity; the trusted execution environment is used to perform real-time de-identification processing on the data to be de-identified within the trusted execution environment based on the hardware root of trust. The traceability and control layer includes a full-chain traceability and data collection module, a trusted storage module, and an integrated control module; The end-to-end traceability and data collection module is used to collect operational behavior information and data flow information at key nodes in the data introduction, dynamic desensitization and data circulation process, and generate traceability records. The trusted storage module is used to receive traceability records after hashing and store them in a trusted manner; the integrated management and control module is used to read traceability records from the trusted storage module for security auditing and risk identification, and dynamically adjust the preset policy rules when the risk score exceeds the preset threshold, and feed the policy adjustment results back to the dynamic desensitization policy engine to form an integrated management and control closed loop.