Public data platform compliance early warning system based on legal risk indicators

Abstract

The invention discloses a common data platform compliance early warning system based on legal risk indicators, which belongs to the technical field of compliance management, and specifically comprises the following steps: a request analysis module receives a data processing request containing a data subject identifier, a processing purpose and a timestamp; the feature extraction module obtains dynamic authorization features such as a historical authorization range sequence and recent change records; the compliance analysis module generates legal risk indexes based on the legal rule base; the decision fusion module generates a compliance early warning decision through multi-dimensional fusion calculation; the execution control module realizes complete release, condition limitation or complete blocking according to the decision; the traceability recording module constructs a complete decision traceability record; according to the invention, through dynamic authorization analysis, multi-level law matching and intelligent decision fusion, accurate early warning and effective management and control of the compliance risk of the public data platform are realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of compliance management, and in particular to a public data platform compliance early warning system based on legal risk indicators. BACKGROUND

[0002] With the rapid development of the digital economy, public data platforms have become important infrastructure for data circulation and value mining. Various institutions carry out data collection, processing and analysis activities through data platforms, promoting the market allocation of data elements. In this context, data compliance management faces unprecedented challenges, especially in complex scenarios such as cross-border data flow and multi-party data collaboration, traditional compliance management methods have been difficult to meet actual needs.

[0003] Currently, public data platforms mainly use compliance checking methods based on fixed rules. These methods match and check data processing requests through a pre-set legal rule library, and trigger an early warning when detecting operations that match the prohibitive provisions in the rule library. Some systems introduce an authorization management mechanism to assist in determining the compliance of data processing requests by recording the authorization status of users. Existing technologies also use rule engines to perform logical reasoning, converting legal provisions into executable judgment rules to achieve automatic identification of some compliance risks.

[0004] However, with the increasing complexity of data application scenarios, single-dimensional authorization judgment has been unable to meet actual needs. In particular, in the face of frequent changes in data processing request patterns, existing technologies lack the ability to continuously learn and analyze user operation characteristics. The authorization preferences and behavior habits of data subjects have obvious temporal characteristics, and traditional systems often ignore these dynamic changes in authorization characteristics. At the same time, existing solutions have limited detection capabilities for abnormal operation behavior, usually based on simple threshold rules, making it difficult to identify complex abnormal patterns. This technical defect results in a lack of flexibility in data access control, which performs poorly in response to new data processing scenarios. SUMMARY

[0005] The purpose of the present application is to provide a public data platform compliance early warning system based on legal risk indicators, which solves the following technical problems: Existing compliance early warning systems analyze authorization status at the static level, lack deep mining of dynamic authorization characteristics, and fixed rules and simple thresholds are difficult to effectively identify complex behavior patterns.

[0006] The purpose of the present application can be achieved by the following technical solutions: The public data platform compliance early warning system based on legal risk indicators comprises: A request analysis module is configured to receive a data processing request of a public data platform, the data processing request comprising a data subject identifier, a data processing purpose description, and a data processing timestamp; A feature extraction module is configured to obtain dynamic authorization features of the data subject, the dynamic authorization features comprising a historical authorization range sequence and a recent authorization change record; A compliance analysis module is configured to generate a legal element set based on a legal rule library, and perform compliance analysis on the data processing request to generate a legal risk indicator; A decision fusion module is configured to perform multi-dimensional fusion calculation on the dynamic authorization features and the legal risk indicator to generate a compliance warning decision; An execution control module is configured to control an execution process of the data processing request according to the compliance warning decision, the execution process comprising complete release, conditional restriction, or complete blocking; A traceability record module is configured to record the basis for generating the compliance warning decision and the execution track of the data processing request, and construct a complete decision traceability record.

[0007] As a further scheme of the present application, in the compliance analysis module, the process of generating the legal risk indicator is as follows: A legal clause set related to the data processing purpose is extracted from the legal rule library, obligation provisions and prohibition provisions in the legal clause set are analyzed, a legal element set is generated, and an operation feature sequence in the data processing request is analyzed, the operation feature sequence comprising a data processing frequency feature, a data processing scale feature, and a data processing object type feature; Multi-level matching calculation is performed on the operation feature sequence and the legal element set, the multi-level matching calculation comprising literal matching calculation and semantic matching calculation, a legal element weight set is generated based on the multi-level matching calculation result, and the legal risk indicator is generated in combination with the legal element weight set and the operation feature sequence.

[0008] As a further scheme of the present application, the process of the multi-level matching calculation is as follows: An association graph of the legal element set and the operation feature sequence is constructed, the association graph comprising legal element nodes and operation feature nodes, direct association strength between the legal element nodes and the operation feature nodes is calculated, the direct association strength being calculated based on a string similarity algorithm, and logical dependency relationships between the legal element nodes are analyzed, the logical dependency relationships comprising prerequisite conditions and mutual exclusion conditions; Based on logical dependencies, the indirect correlation strength between legal element nodes and operational feature nodes is derived. The direct and indirect correlation strengths are weighted and fused to generate a comprehensive correlation value. The legal element nodes are sorted according to the comprehensive correlation value to generate a legal element weight set. The legal element weight set and the operational feature sequence are subjected to matrix operation to generate the initial value of the legal risk indicator. The initial value of the legal risk indicator is standardized to obtain the final legal risk indicator.

[0009] As a further aspect of the present invention: in the decision fusion module, the multi-dimensional fusion calculation process is as follows: Obtain time-series data of dynamic authorization characteristics, including historical authorization change time points and authorization scope change trajectories; obtain real-time calculated values ​​of legal risk indicators; perform trend analysis on the time-series data of dynamic authorization characteristics to generate authorization stability assessment values; and perform fluctuation analysis on the real-time calculated values ​​of legal risk indicators to generate risk fluctuation coefficients. A dynamic weight adjustment model is established. The dynamic weight adjustment model adjusts the weight ratio of dynamic authorization features and legal risk indicators based on the authorization stability assessment value and risk volatility coefficient. The adjusted weight ratio is used to perform weighted calculation on the standardized dynamic authorization features and normalized legal risk indicators to generate an early warning decision score. The corresponding compliance early warning decision is generated based on the threshold range of the early warning decision score.

[0010] As a further aspect of the present invention: the construction process of the dynamic weight adjustment model is as follows: Obtain the authorized stability assessment value sequence and risk volatility coefficient sequence from historical early warning decision records, calculate the first correlation coefficient between the authorized stability assessment value sequence and the early warning decision accuracy rate, calculate the second correlation coefficient between the risk volatility coefficient sequence and the early warning decision accuracy rate, and determine the degree of influence of the authorized stability assessment value and risk volatility coefficient in weight adjustment based on the correlation coefficients; A weight adjustment function is established. The input of the weight adjustment function includes the current authorization stability assessment value and the current risk volatility coefficient. The weight adjustment function outputs the weight ratio of dynamic authorization characteristics and legal risk indicators. The weight adjustment function adopts a linear transformation method. The parameters of the linear transformation are dynamically adjusted according to two correlation coefficients. The weight adjustment function is integrated into the dynamic weight adjustment model. The dynamic weight adjustment model updates the parameters of the weight adjustment function once at fixed time intervals.

[0011] As a further aspect of the present invention: in the execution control module, the execution process of the control data processing request is as follows: Analyze the decision type and restriction level in compliance early warning decision-making, select the execution path according to the decision type, and the execution path includes a complete release path, a conditional restriction path, and a complete blocking path. For the conditional restriction path, generate a dynamic data masking scheme based on the restriction level. The dynamic data masking scheme includes field-level masking rules and record-level masking rules. The dynamic data masking scheme is compiled into database query and modification instructions, which are injected during the execution of data processing requests to perform real-time masking processing on the output results. The actual number of masked fields and the proportion of masked records are recorded during the masking process.

[0012] As a further aspect of the present invention: the process of generating the dynamic data masking scheme is as follows: Obtain the set of data fields and the set of data records involved in the data processing request; calculate the occlusion strength value of each data field based on legal risk indicators; calculate the visibility score of each data record based on dynamic authorization features; and establish a data field occlusion priority sequence and a data record visibility sequence. Based on the restriction level, determine the occlusion threshold and visibility threshold, generate a set of field-level occlusion rules and a set of record-level occlusion rules. The set of field-level occlusion rules specifies the occlusion method and degree of each data field, and the set of record-level occlusion rules specifies the visibility range and access permissions of each data record. Perform consistency verification on the set of field-level occlusion rules and the set of record-level occlusion rules, and output a complete dynamic data occlusion scheme.

[0013] As a further aspect of the present invention: in the source tracing recording module, the process of constructing a complete decision source tracing record is as follows: The system collects timestamps and operation parameters of key operation nodes, including request receiving nodes, feature query nodes, indicator calculation nodes, decision generation nodes, and decision execution nodes. The collected data is combined into an event record sequence according to time order. The event record sequence is hashed to generate a sequence integrity verification value. The event record sequence and the sequence integrity verification value are stored in the blockchain network. Blocks containing timestamps are generated in the blockchain network, and an association index between the event record sequence and the original data processing request is established.

[0014] The beneficial effects of this invention are: This invention continuously tracks the authorization trends of data subjects through a dynamic authorization feature analysis module, and achieves deep correlation analysis between legal clauses and operational characteristics by combining a multi-level legal element matching method. It utilizes a dynamic weight adjustment model to adaptively adjust evaluation weights based on authorization stability and risk fluctuations, overcoming the rigidity of fixed-rule systems. Real-time data masking technology enables precise control over data processing requests at different risk levels, ensuring compliance while maintaining data utility. A blockchain-based decision tracing mechanism fully records the entire process from request reception to decision execution, forming an immutable audit chain. The synergistic effect of these technical features enables the system to accurately identify complex abnormal behavior patterns, dynamically adapt to changing data processing scenarios, effectively improve the accuracy and adaptability of compliance warnings, and address the technical shortcomings of traditional systems in dynamic authorization analysis, complex risk identification, and multi-dimensional feature fusion. Attached Figure Description

[0015] The invention will now be further described with reference to the accompanying drawings.

[0016] Figure 1 This is a schematic diagram of the modules of the present invention. Detailed Implementation

[0017] 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.

[0018] Please see Figure 1 As shown, this invention is a compliance early warning system for public data platforms based on legal risk indicators, comprising: The request parsing module is used to receive data processing requests from the public data platform. The data processing request includes the data subject identifier, a description of the data processing purpose, and a data processing timestamp. The feature extraction module is used to obtain the dynamic authorization features of the data subject. The dynamic authorization features include the historical authorization range sequence and the recent authorization change record. The compliance analysis module generates a set of legal elements based on a legal rule base, performs compliance analysis on data processing requests, and generates legal risk indicators. The decision fusion module is used to perform multi-dimensional fusion calculations of dynamic authorization features and legal risk indicators to generate compliance early warning decisions; The execution control module is used to control the execution process of data processing requests based on compliance warning decisions. The execution process includes complete permission, conditional restriction, or complete blocking. The traceability record module is used to record the basis for the generation of compliance warning decisions and the execution trajectory of data processing requests, thus building a complete decision traceability record.

[0019] In the request parsing module of this invention, upon receiving a data processing request, the format of the data subject identifier is first validated. By comparing it with the unique identifier system in the subject information database, the validity of the identifier is confirmed and associated with the corresponding subject file, ensuring data traceability. For the description of the data processing purpose, natural language processing technology is used for structured transformation, extracting core operation types and data usage keywords, mapping them to a preset processing scenario classification system to clarify the scope of the operation. The data processing timestamp is synchronized with the system's reference clock for calibration, generating a standardized time record with time zone information to ensure consistency in time series analysis.

[0020] In the feature extraction module of this invention, historical authorization data of the data subject is retrieved in batches and sorted by timestamp to form a historical authorization range sequence. The sequence includes the data field set, usage scenario restrictions, and permission validity period of authorizations for each time period. At the same time, authorization change events are monitored in real time, capturing recent authorization addition, modification, and revocation operations, recording the change trigger time, operation subject, and specific scope adjustment content, and performing time-series alignment and structured integration on the two types of data to form a dynamically updated authorization feature dataset.

[0021] In the compliance analysis module of this invention, the process of generating legal risk indicators is as follows: First, a set of related legal clauses is extracted from a structured legal rule base. The legal rule base is pre-categorized and stored according to data processing scenarios, with each scenario tag corresponding to multiple sets of related legal clauses. The module precisely matches the core keywords in the description of the data processing purpose with the scenario tags in the rule base to filter out a set of legal clauses directly related to the current processing behavior. If cross-scenario related clauses exist, the system will expand the extraction through the correlation index between clauses to ensure complete clause coverage.

[0022] After extracting the set of clauses, the module activates the legal clause parsing engine to perform semantic breakdown and type determination on each clause. Using named entity recognition and relation extraction algorithms in natural language processing, it identifies the obligatory and prohibitive provisions within the clauses. The identification of obligatory provisions focuses on the behavioral requirements corresponding to mandatory expressions such as "shall" and "must," while prohibitive provisions target the behavioral restrictions corresponding to negative expressions such as "shall not" and "prohibited." Based on the identification results, the module further extracts core elements from the clauses, including data processing entity qualification requirements, data scope limitations, and operational procedure specifications. These elements are then structured and integrated into a legal element set, with each element containing element type, binding content, and associated clause ID.

[0023] While constructing the legal element set, the module simultaneously analyzes the operational characteristic sequence in data processing requests. For data processing frequency characteristics, the module connects to the request log system to count the number of similar data processing requests initiated by the same entity per unit time, forming a frequency distribution characteristic over time. For data processing scale characteristics, the module determines the number of data records or data file bytes processed per instance by parsing the data query range and data volume identifier in the request parameters. For data processing object type characteristics, the module retrieves classification and grading information from the data asset catalog to determine the sensitivity level and data type of the processed data, such as personal identification information or trade secret data. These three types of characteristics are combined according to a preset format to form a complete operational characteristic sequence.

[0024] The process of multi-level matching calculation is as follows: The multi-level matching calculation begins with a set of legal elements and a sequence of operational features as input, first constructing a relational graph. The relational graph uses a graph database storage architecture, where legal element nodes contain attributes such as element ID, element type, and constraint content, while operational feature nodes contain attributes such as feature type, feature value, and data source. The module achieves initial mapping by establishing relational edges between nodes. The initial attributes of these edges only mark the basic relational relationship and do not include strength information.

[0025] After the graph construction is completed, the module calculates the direct association strength between legal element nodes and operational feature nodes. A string similarity algorithm is used to compare the element constraint content with the feature description text. Initial similarity values ​​are obtained by calculating the overlap of keywords in the text and the cosine similarity of semantic vectors. For numerical features such as processing scale, the module performs interval matching between the scale limit threshold in the legal element and the actual processing scale. The similarity values ​​are adjusted based on the deviation between the actual scale and the threshold, ultimately generating the direct association strength between each node.

[0026] After calculating the direct association strength, the module analyzes the logical dependencies between legal element nodes. By traversing the associated clause IDs of each element in the legal element set, it retrieves the logical expressions in the original clause text to identify the preconditions and mutually exclusive conditions between elements. Precondition relationships are determined by logical expressions such as "only if element A is satisfied can the operation corresponding to element B be executed," while mutually exclusive conditions are identified based on the expression "if element A is satisfied, element B cannot be satisfied." The module adds the identified logical relationships to the association graph as directed edges, with edge attributes labeled with dependency type and dependency strength.

[0027] Based on logical dependencies, the module derives the strength of indirect associations between legal element nodes and operational feature nodes. If legal element A and element B have a preconditional dependency, and element B and operational feature C have a direct association strength, then element A and feature C form an indirect association. The strength of this indirect association is the product of the dependency strength between element A and B and the direct association strength between element B and C. For node pairs with multiple indirect associations, the module takes the maximum value of the association strengths along each path as the final indirect association strength.

[0028] Subsequently, the module performs a weighted fusion of direct and indirect correlation strengths. The weight allocation is determined based on the importance level of the legal elements; elements corresponding to prohibitive provisions have higher weights than those corresponding to obligatory provisions, and elements involving core sensitive data have higher weights than ordinary data elements. The two types of strengths are then weighted and summed to obtain the comprehensive correlation value between each legal element node and the operational feature sequence.

[0029] Based on the comprehensive correlation scores, the module sorts the legal element nodes in descending order, with elements having higher correlation scores receiving greater weight in subsequent calculations. The sorted elements are then assigned weights according to their correlation scores, forming a legal element weight set. The module uses matrix operations to multiply the numerical vector of the operational feature sequence with the weight vector of the legal element weight set to obtain the initial value of the legal risk indicator. To eliminate the influence of differences in the numerical ranges of different feature dimensions, the module employs a min-max standardization method, mapping the initial value to a standardized range of 0-1, ultimately generating a standardized legal risk indicator.

[0030] In the decision fusion module of this invention, the process of multi-dimensional fusion calculation is as follows: The module uses dynamic authorization feature data and legal risk indicator data as core inputs. First, it interfaces with the feature extraction module's time-series database to obtain the dynamic authorization feature time-series data of the target data subject. This time-series data includes historical authorization change time points and authorization scope change trajectories. The historical authorization change time points precisely record the system time of each authorization addition, modification, and revocation. The authorization scope change trajectory details the data field access permissions, processing scenario restrictions, and authorization validity period adjustments before and after each change. The data is arranged in ascending order of timestamps to form a structured time-series dataset. Simultaneously, the module receives real-time calculated values ​​of legal risk indicators output by the compliance analysis module through a real-time data synchronization mechanism. These values ​​have been standardized, with their ranges mapped to preset intervals to ensure compatibility with subsequent data processing.

[0031] For time-series data with dynamic authorization characteristics, the module employs a sliding window analysis method for trend analysis. By setting a fixed-duration analysis window, the frequency of authorization changes and the magnitude of authorization range adjustments within the window are statistically analyzed. Combined with time-series smoothing techniques, short-term fluctuations are eliminated to identify long-term trends in authorization changes. If the number of authorization changes and the magnitude of range adjustments within the window are low, the authorization status is considered stable, generating a high authorization stability assessment value. Conversely, if changes are frequent and the range fluctuations are large, the authorization status is considered unstable, generating a low authorization stability assessment value. The assessment value is ultimately presented through numerical quantification, and the value is positively correlated with stability.

[0032] In the legal risk indicator volatility analysis phase, the module selects real-time calculated values ​​of legal risk indicators from multiple recent consecutive time points to construct a short-term numerical sequence. The volatility of this sequence is calculated using analysis of variance (ANOVA) to quantify the dispersion of the indicator values. If the numerical sequence is generally stable with low dispersion, a smaller risk volatility coefficient is generated; if the values ​​experience significant fluctuations in the short term with high dispersion, a larger risk volatility coefficient is generated. The coefficient value is positively correlated with the volatility level, reflecting the current dynamic trend of legal risk.

[0033] After the dynamic weight adjustment model is activated, it first reads the current authorization stability assessment value and risk volatility coefficient, and then adjusts the weight ratio of dynamic authorization features and legal risk indicators according to preset mapping rules. When the authorization stability assessment value is high, it indicates that the data subject has clear authorization preferences, and the reference value of dynamic authorization features for decision-making is enhanced, so the corresponding weight ratio is increased. When the risk volatility coefficient is high, it indicates that the current legal risk is highly uncertain, and the decision-making influence of legal risk indicators needs to be strengthened, so the corresponding weight ratio is increased.

[0034] The module standardizes dynamic authorization features by mapping numerical values ​​of different dimensions to a unified range using a data normalization algorithm, eliminating dimensional differences. It also normalizes the real-time calculated values ​​of legal risk indicators to ensure consistency with the numerical magnitudes of dynamic authorization features. Using adjusted weight ratios, the two types of standardized data are weighted separately, and the sum of the weighted results yields the early warning decision score. The module has built-in preset threshold ranges, with different ranges corresponding to different compliance early warning decision types. Based on the specific range the early warning decision score falls into, it automatically generates compliance early warning decisions of complete approval, conditional restrictions, or complete blocking.

[0035] The construction process of the dynamic weight adjustment model is as follows: By connecting the source tracing record module with the decision log database, key data is extracted from historical early warning decision records, including the authorized stability assessment value, risk volatility coefficient, and the early warning decision accuracy rate obtained from subsequent verification of each decision. The authorized stability assessment values ​​are organized into an authorized stability assessment value sequence in chronological order, and the risk volatility coefficients are organized into a risk volatility coefficient sequence in the corresponding chronological order, ensuring that the two types of sequences correspond one-to-one with the early warning decision accuracy data, forming a complete historical sample dataset.

[0036] The module uses statistical analysis methods to calculate correlation coefficients. For the authorized stability assessment value series and the early warning decision accuracy rate series, a correlation analysis algorithm is used to calculate the linear correlation between the two, yielding the first correlation coefficient. The same algorithm is used to calculate the second correlation coefficient for the risk volatility coefficient series and the early warning decision accuracy rate series. A larger absolute value of the correlation coefficient indicates a stronger correlation between the corresponding indicator and the early warning decision accuracy rate, and a higher degree of influence in weight adjustment. If the correlation coefficient is positive, it indicates a positive correlation between the indicator value and the decision accuracy rate, and the indicator weight should be increased as the value increases.

[0037] Based on the calculation results of the first and second correlation coefficients, the module determines the weights of the authorization stability assessment value and the risk volatility coefficient in the weight adjustment. Indicators with higher correlation correspond to higher influence weights, which will serve as the basis for subsequent calibration of the weight adjustment function parameters. Subsequently, the module constructs a weight adjustment function that takes the current authorization stability assessment value and the current risk volatility coefficient as input parameters and outputs the weight ratio of dynamic authorization characteristics and legal risk indicators. The function employs a linear transformation, mapping the input and output by setting linear parameters such as slope and intercept. The parameters of the linear transformation are dynamically adjusted according to the first and second correlation coefficients, with indicators with higher correlation corresponding to larger parameter weights, ensuring that the function output reflects the actual impact of the indicators on decision-making.

[0038] The module integrates the constructed weight adjustment function into the core algorithm layer of the dynamic weight adjustment model, and sets up a parameter update mechanism. At fixed time intervals, the model automatically retrieves newly added historical early warning decision records within that period, recalculates the first and second correlation coefficients, and calibrates the linear parameters of the weight adjustment function based on the new correlation coefficient values, achieving dynamic optimization of the function parameters. The parameter update cycle can be preset according to the system's data volume and business scenario requirements, ensuring that the model can adapt to changes in data processing behavior patterns and legal rules, maintaining the accuracy of weight adjustment.

[0039] In the execution control module of this invention, the execution process of the control data processing request is as follows: Upon receiving the compliance warning decision from the decision fusion module, the decision parsing engine is first activated. This engine connects to the decision output end through a standardized interface, extracts the structured data from the compliance warning decision, and clarifies the decision type and restriction level. The decision type field directly identifies three execution directions: complete approval, conditional restriction, or complete blocking. The restriction level field quantifies the strictness of the conditional restrictions according to preset levels (such as high, medium, and low). The parsing results are stored in the module's local cache to ensure rapid retrieval in subsequent processes.

[0040] Based on the decision type obtained from the parsing, the module automatically selects the corresponding execution path. Under the fully open path, the module does not intervene in the data processing request, directly forwards the original request to the core processing node of the public data platform, and simultaneously records the request forwarding time and the target node identifier; under the fully blocked path, the module immediately terminates the request process, generates a structured response message containing the blocking reason (such as violation of prohibitive legal provisions, mismatch of authorization scope), feeds it back to the request initiator, and writes the blocking event to the system exception log.

[0041] For conditionally restricted paths, the module triggers a dynamic data masking scheme generation process, with the generation logic strongly correlated with the restriction level. At a high restriction level, the scheme needs to cover more sensitive fields and unauthorized records; at a low restriction level, the scheme only masks core sensitive data. After the scheme is generated, the module calls the masking rule compilation component to convert the field-level and record-level rules in the scheme into database-executable query modification instructions. For SQL-type data requests, the component uses syntax parsing technology to identify the SELECT field and WHERE condition in the query statement, embedding the masking rules into field projection and row filtering logic to generate modified SQL instructions; for unstructured data requests, the component generates binary processing instructions for field replacement or record filtering.

[0042] During the execution of data processing requests, the module injects modified instructions into the database before the query command is sent through a request interception mechanism. After the database executes the injected instructions, the returned result set already contains the real-time masking effect. The module synchronously starts a masking effect statistics component, traverses the processed result set, and counts the actual number of masked fields (e.g., the total number of fields masked in each record) and the proportion of masked records (the proportion of filtered or partially masked records to the total number of records). After associating the statistical data with the request ID and processing timestamp, the data is stored in the masking audit database for subsequent compliance traceability.

[0043] The process of generating the dynamic data masking scheme is as follows: The first step in generating a dynamic data masking solution is to obtain the target data set. The module connects to the data catalog service of a public data platform, parses resource identifiers (such as database table names and file IDs) in the data processing request, retrieves the corresponding data structure metadata, and determines the set of data fields involved in the request (including field names, data types, and sensitivity level tags). Simultaneously, the module queries the current total number and distribution of data records for the resource through a data access interface, forming the basic information of the data record set, including record ID ranges and associated data subject identifiers, ensuring that the solution covers all data dimensions involved in the request.

[0044] Based on the legal risk indicators output by the compliance analysis module, the module calculates the masking strength value for each data field. Constraints directly related to the field within the legal risk indicators (such as a field needing to meet "de-identified storage" requirements) have higher weights. The module quantifies and generates masking strength values ​​based on the mapping relationship between indicator values ​​and field sensitivity levels. A higher strength value means that the field requires more stringent masking processing (such as full-field masking rather than partial masking). Simultaneously, the module combines dynamic authorization features provided by the feature extraction module to calculate the visibility score for each data record. If the historical authorization scope in the authorization features includes the record's associated entity identifier or data purpose, the score increases accordingly; if there is a recent authorization revocation record, the score decreases. The score directly reflects the record's accessibility to the current request.

[0045] The module sorts the data field set in descending order based on the occlusion intensity value, forming a data field occlusion priority sequence. Fields with higher priority are included in the occlusion scope first in the scheme. It also sorts the data record set in descending order based on visibility score, forming a data record visibility sequence. Records with lower scores are included in the restricted access scope first. After sorting, the module determines the occlusion threshold and visibility threshold based on the restriction level of the conditional restriction path. Higher restriction levels correspond to lower occlusion thresholds (i.e., fields with intensity values ​​exceeding the lower threshold must be occluded) and lower visibility thresholds (i.e., records with scores below the lower threshold must be restricted); lower restriction levels correspond to higher thresholds, only occluding high-risk fields and low-visibility records.

[0046] Based on the aforementioned thresholds, the module generates field-level and record-level masking rule sets. For each field to be masked, the field-level masking rule set specifies the masking method (e.g., partial masking, replacement with placeholders, format conversion) and the degree of masking (e.g., masking 80% or 50% of the field length), and associates it with the corresponding legal basis (e.g., a rule corresponds to a specific clause of the Data Security Law). For each record to be restricted, the record-level masking rule set specifies the visibility range (e.g., only allowing viewing of non-sensitive fields in the record) and access permissions (e.g., prohibiting downloads, allowing only read-only access). After rule generation, the module starts a consistency verification engine to check whether field-level rules conflict with data types (e.g., numeric fields are not suitable for string replacement masking) and whether record-level rules contradict authorization characteristics (e.g., authorized records are mistakenly judged as restricted access). After successful verification, a complete dynamic data masking scheme containing the rule set, threshold parameters, and verification results is output.

[0047] In the source tracing recording module of this invention, the process of constructing a complete decision source tracing record is as follows: Real-time collection of timestamps and operation parameters for key operation nodes. The request receiving node needs to collect the unique request ID, data subject identifier, request initiator IP, and request message digest; the feature query node collects the query subject identifier, query time, and the returned dynamic authorization feature data digest; the indicator calculation node collects the input operation feature sequence, legal element set, and output legal risk indicator value; the decision generation node collects the authorization stability assessment value, risk fluctuation coefficient, and final early warning decision result; the decision execution node collects the execution path type, masking rule ID, and a snapshot of the actual execution result.

[0048] After data collection, the module sorts the data in ascending order by the timestamps of each node's operation. It then combines the collected timestamps, node identifiers, and operation parameter snapshots into a structured event record sequence. Each record in the sequence contains a globally unique event ID to ensure traceability. The module uses a hash algorithm to perform integrity calculations on the event record sequence, generating a fixed-length sequence integrity check value. This value is used to subsequently verify whether the sequence has been tampered with.

[0049] The module then uses a blockchain adaptation interface to write the event record sequence and its corresponding integrity verification value into a pre-defined blockchain network. The blockchain network employs a consortium blockchain architecture, and the writing process requires consensus verification from all nodes. Once verified, a new block is generated containing the current timestamp, the event record sequence hash, and the hash of the previous block. This new block is synchronized to all nodes in the blockchain network to ensure the data is immutable.

[0050] Finally, the module establishes an index linking the event record sequence with the original data processing request. The index uses the unique ID of the original request as the primary key, and associates the event record sequence with the block height and storage address in the blockchain. It also records the index creation timestamp, which facilitates quick location and retrieval of the complete decision traceability record through the request ID, enabling full-process auditability.

[0051] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A public data platform compliance early warning system based on legal risk indicators, characterized in that: include: The request parsing module is used to receive data processing requests from the public data platform. The data processing request includes the data subject identifier, a description of the data processing purpose, and a data processing timestamp. The feature extraction module is used to obtain the dynamic authorization features of the data subject. The dynamic authorization features include the historical authorization range sequence and the recent authorization change record. The compliance analysis module generates a set of legal elements based on a legal rule base, performs compliance analysis on data processing requests, and generates legal risk indicators. The decision fusion module is used to perform multi-dimensional fusion calculations of dynamic authorization features and legal risk indicators to generate compliance early warning decisions; The execution control module is used to control the execution process of data processing requests based on compliance warning decisions. The execution process includes complete permission, conditional restriction, or complete blocking. The traceability record module is used to record the basis for the generation of compliance warning decisions and the execution trajectory of data processing requests, thus building a complete decision traceability record.

2. The public data platform compliance early warning system based on legal risk indicators according to claim 1, characterized in that, In the compliance analysis module, the process of generating legal risk indicators is as follows: Extract a set of legal clauses related to the data processing purpose from the legal rule base, analyze the obligatory and prohibitive provisions in the legal clause set, and generate a set of legal elements; analyze the operational feature sequence in the data processing request, which includes data processing frequency characteristics, data processing scale characteristics, and data processing object type characteristics; The operation feature sequence is matched with the legal element set at multiple levels. The multiple level matching calculation includes literal matching calculation and semantic matching calculation. Based on the results of the multiple level matching calculation, a legal element weight set is generated. The legal element weight set and the operation feature sequence are combined to generate a legal risk indicator.

3. The public data platform compliance early warning system based on legal risk indicators according to claim 2, characterized in that, The process of multi-level matching calculation is as follows: Construct a relationship graph between a set of legal elements and a sequence of operational features. The relationship graph contains legal element nodes and operational feature nodes. Calculate the direct relationship strength between legal element nodes and operational feature nodes. The direct relationship strength is calculated based on a string similarity algorithm. Analyze the logical dependencies between legal element nodes. The logical dependencies include preconditions and mutually exclusive conditions. Based on logical dependencies, the indirect correlation strength between legal element nodes and operational feature nodes is derived. The direct and indirect correlation strengths are weighted and fused to generate a comprehensive correlation value. The legal element nodes are sorted according to the comprehensive correlation value to generate a legal element weight set. The legal element weight set and the operational feature sequence are subjected to matrix operation to generate the initial value of the legal risk indicator. The initial value of the legal risk indicator is standardized to obtain the final legal risk indicator.

4. The public data platform compliance early warning system based on legal risk indicators according to claim 1, characterized in that, In the decision fusion module, the multi-dimensional fusion calculation process is as follows: Obtain time-series data of dynamic authorization characteristics, including historical authorization change time points and authorization scope change trajectories; obtain real-time calculated values ​​of legal risk indicators; perform trend analysis on the time-series data of dynamic authorization characteristics to generate authorization stability assessment values; and perform fluctuation analysis on the real-time calculated values ​​of legal risk indicators to generate risk fluctuation coefficients. A dynamic weight adjustment model is established. The dynamic weight adjustment model adjusts the weight ratio of dynamic authorization features and legal risk indicators based on the authorization stability assessment value and risk volatility coefficient. The adjusted weight ratio is used to perform weighted calculation on the standardized dynamic authorization features and normalized legal risk indicators to generate an early warning decision score. The corresponding compliance early warning decision is generated based on the threshold range of the early warning decision score.

5. The public data platform compliance early warning system based on legal risk indicators according to claim 4, characterized in that, The construction process of the dynamic weight adjustment model is as follows: Obtain the authorized stability assessment value sequence and risk volatility coefficient sequence from historical early warning decision records, calculate the first correlation coefficient between the authorized stability assessment value sequence and the early warning decision accuracy rate, calculate the second correlation coefficient between the risk volatility coefficient sequence and the early warning decision accuracy rate, and determine the degree of influence of the authorized stability assessment value and risk volatility coefficient in weight adjustment based on the correlation coefficients; A weight adjustment function is established. The input of the weight adjustment function includes the current authorization stability assessment value and the current risk volatility coefficient. The weight adjustment function outputs the weight ratio of dynamic authorization characteristics and legal risk indicators. The weight adjustment function adopts a linear transformation method. The parameters of the linear transformation are dynamically adjusted according to two correlation coefficients. The weight adjustment function is integrated into the dynamic weight adjustment model. The dynamic weight adjustment model updates the parameters of the weight adjustment function once at fixed time intervals.

6. The public data platform compliance early warning system based on legal risk indicators according to claim 1, characterized in that, In the execution control module, the execution process of the control data processing request is as follows: Analyze the decision type and restriction level in compliance early warning decision-making, select the execution path according to the decision type, and the execution path includes a complete release path, a conditional restriction path, and a complete blocking path. For the conditional restriction path, generate a dynamic data masking scheme based on the restriction level. The dynamic data masking scheme includes field-level masking rules and record-level masking rules. The dynamic data masking scheme is compiled into database query and modification instructions, which are injected during the execution of data processing requests to perform real-time masking processing on the output results. The actual number of masked fields and the proportion of masked records are recorded during the masking process.

7. The public data platform compliance early warning system based on legal risk indicators according to claim 6, characterized in that, The process of generating the dynamic data masking scheme is as follows: Obtain the set of data fields and the set of data records involved in the data processing request; calculate the occlusion strength value of each data field based on legal risk indicators; calculate the visibility score of each data record based on dynamic authorization features; and establish a data field occlusion priority sequence and a data record visibility sequence. Based on the restriction level, determine the occlusion threshold and visibility threshold, generate a set of field-level occlusion rules and a set of record-level occlusion rules. The set of field-level occlusion rules specifies the occlusion method and degree of each data field, and the set of record-level occlusion rules specifies the visibility range and access permissions of each data record. Perform consistency verification on the set of field-level occlusion rules and the set of record-level occlusion rules, and output a complete dynamic data occlusion scheme.

8. The public data platform compliance early warning system based on legal risk indicators according to claim 1, characterized in that, In the source tracing record module, the process of constructing a complete decision source tracing record is as follows: The system collects timestamps and operation parameters of key operation nodes, including request receiving nodes, feature query nodes, indicator calculation nodes, decision generation nodes, and decision execution nodes. The collected data is combined into an event record sequence according to time order. The event record sequence is hashed to generate a sequence integrity verification value. The event record sequence and the sequence integrity verification value are stored in the blockchain network. Blocks containing timestamps are generated in the blockchain network, and an association index between the event record sequence and the original data processing request is established.

Images