Data quality assessment method for trusted data space nodes

Abstract

The invention discloses a data quality evaluation method for trusted data space nodes, which comprises the following steps of: extracting metadata of a database by each node through a locally deployed data quality evaluation service, the metadata comprises at least one of data structure information, field types, data formats, constraint conditions, data volume statistical information and updating time, and does not contain any original data content; the node encrypts the metadata and uploads the metadata to a central control board through a secure communication channel; the central control board receives and decrypts the metadata, analyzes data features of the metadata and matches the metadata with a preset evaluation rule base; the principle of'visible and unavailable 'of a trusted data space is perfectly met, all operations related to original data are completed locally in a data source, the risk of original data leakage is fundamentally eradicated, and meanwhile the data quality can be accurately evaluated.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data quality evaluation, and particularly relates to a data quality evaluation method for a trusted data space node. BACKGROUND

[0002] With the acceleration of the global digitalization process, data has become a key production factor. The demand for data sharing and collaboration across organizations and regions is increasingly urgent. As a new data exchange infrastructure, trusted data space provides a possibility for data providers to share data under the premise of safeguarding data sovereignty and security. Its core principle is "visible but not available", that is, participants can know what data exists in the space and its basic characteristics (through metadata), but cannot directly access the original data content of others. The original data is always kept in the local environment of the data provider.

[0003] However, this advanced data security paradigm also brings great challenges to traditional data quality management. Data quality is the cornerstone of data value release, and usually needs to be evaluated in multiple dimensions such as completeness, accuracy, consistency, and timeliness. Traditional data quality evaluation methods rely heavily on direct access and analysis of the original data, usually requiring data to be centralized in a unified data platform or quality tool, and running predefined rules and algorithms for detection.

[0004] Under the constraints of trusted data space, this centralized processing mode becomes unfeasible. The evaluation party cannot access the original data, which makes it impossible to directly judge whether the data provided by a certain node in the space is real and reliable, and whether it meets the expected quality requirements. For example, in the joint modeling scenario of financial risk control, multiple banks want to share customer feature data, but cannot verify whether there are a large number of null values or incorrect labels in the feature data provided by the other party; in the medical research scenario, research institutions cannot confirm the format consistency and logical accuracy of the anonymized medical record data provided by multiple hospitals. This "quality black box" state seriously hinders the establishment of trust between data space participants and the value mining, and limits the wide application and long-term development of data space.

[0005] Currently, the industry and academia have proposed some technologies for data calculation under privacy protection, such as secure multi-party computation (MPC) and federated learning (FL), but they mainly focus on joint modeling or statistical analysis in ciphertext or decentralized state, and are not specifically designed to solve the systematic data quality evaluation problem, and usually have huge computational overhead, making it difficult to cover the rich rule types involved in data quality evaluation.

[0006] Therefore, there is an urgent need in the art for an innovative technical solution that can adapt to the unique architecture of trusted data space, and achieve effective evaluation and global control of node data quality without touching the original data. SUMMARY

[0007] The present application aims to provide a data quality evaluation method for trusted data space nodes, which takes "metadata-driven, rule-sinking, local calculation, centralized control" as the core idea, and through a set of precise process design and security mechanisms, realizes the standardized, automated and trusted evaluation of the data quality of each node in the space without violating the "visible but not available" principle, to solve the problems raised in the background art.

[0008] To achieve the above object, the present application provides the following technical solution: a data quality evaluation method for trusted data space nodes, comprising the following steps:

[0009] Each node extracts the metadata of its database through the locally deployed data quality evaluation service, the metadata including at least one of data structure information, field type, data format, constraint condition, data volume statistical information and update time, and not containing any original data content;

[0010] The node encrypts the metadata and uploads it to the central console through a secure communication channel;

[0011] The central console receives and decrypts the metadata, analyzes its data characteristics, and matches it with a pre-set evaluation rule library, the matching process being realized through multi-dimensional feature similarity calculation, and if the matching similarity is lower than a pre-set threshold, an artificial rule customization process is started;

[0012] The central console encapsulates and encrypts the evaluation rule configuration file generated by the matching or customization, the configuration file including rule identification, evaluation dimension, evaluation index, rule pre-set value, evaluation logic, execution order and check code, and issues it to the corresponding node through a secure channel;

[0013] The node receives and verifies the configuration file, and performs a data quality evaluation operation locally, analyzes the local data according to the evaluation rules, and generates a data quality evaluation report containing the evaluation score, abnormal records and bias analysis;

[0014] The node encrypts the evaluation report and uploads it to the central console;

[0015] The central console summarizes the evaluation reports of each node, performs a comprehensive data quality analysis, generates a global data quality view, and supports visual display and early warning.

[0016] As a preferred technical solution in the application, the metadata extraction process specifically comprises:

[0017] The local database is connected through a data quality assessment service to read metadata information in a system table or a data dictionary;

[0018] The extracted data table structure information includes a table name, a field name, a data type, a data length, whether it can be empty, a default value, and a primary-foreign key relationship;

[0019] The extracted data statistical information includes a total number of records, a null value proportion, a number of unique values, a numerical distribution interval, and an update frequency;

[0020] The extracted metadata is standardized and converted into a unified JSON or XML format for transmission and parsing.

[0021] As a preferred technical solution in the application, the metadata encryption process adopts an asymmetric encryption algorithm, specifically: the node uses a public key provided by the central console to encrypt the metadata, and the encryption algorithm is RSA or ECC; the central console uses its private key to decrypt the received encrypted metadata; to further improve security, the metadata is compressed before encryption to reduce the amount of transmission data.

[0022] As a preferred technical solution in the application, the multi-dimensional feature similarity calculation in the evaluation rule matching process adopts the following formula:

[0023] S(M,R)=

[0024] Where S(M,R) represents the overall matching similarity of metadata M and rule R, w i represents the weight of the i-th feature item, which is determined by expert experience or a machine learning model according to feature importance, sim(f i M ,f i R ) is the similarity function of feature i, for numerical features (such as data length), the reciprocal of relative error can be used, and for categorical features (such as data type), Jaccard similarity coefficient or cosine similarity can be used for calculation; set the similarity threshold θ=0.85, when S(M,R)≥θ, it is determined that the matching is successful, otherwise the matching fails.

[0025] As a preferred technical solution in the application, the artificial rule customization process comprises:

[0026] The central console pushes the metadata feature information that is not matched successfully to an artificial rule customization platform;

[0027] The platform notifies the technical staff, who can then view the detailed feature information of the metadata.

[0028] Based on data business scenarios, industry standards, and historical experience, technical personnel create or modify evaluation rules on the platform, defining evaluation dimensions, indicators, predetermined values, and logic.

[0029] After the newly customized rules are tested and verified, they are added to the evaluation rule base in the central console and associated with the corresponding metadata features for subsequent automatic matching.

[0030] As a preferred technical solution of the present invention, the encapsulation and encryption processing of the evaluation rule configuration file specifically includes:

[0031] The configuration file is encapsulated using a specific structure, including a rule metadata area and a rule body area;

[0032] The rule metadata area includes the rule ID, version number, applicable data type, creation time, and author;

[0033] The rule body includes the specific evaluation condition expression, threshold parameters, and execution script;

[0034] Use a hash algorithm (such as SHA-256) to generate a digital digest of the complete configuration file content as a checksum;

[0035] The central control console uses a symmetric encryption algorithm (such as AES-256) to generate a random key to encrypt the configuration file, and then uses the public key of the target node to encrypt the random key to form the final encrypted data packet.

[0036] As a preferred technical solution of the present invention, the process of the node receiving and decrypting the configuration file includes: after receiving the encrypted data packet, the node first uses its own private key to decrypt and obtain the symmetric key; uses the symmetric key to decrypt and obtain the plaintext of the configuration file; the node locally calculates the hash value of the configuration file content and compares it with the received check code. If they match, the verification passes and the rule is stored in the local rule base; if they do not match, an error report is sent to the central console and a retransmission is requested.

[0037] As a preferred technical solution of the present invention, the node locally performs the data quality assessment operation, including:

[0038] The data quality assessment service on the node loads assessment rules from the local rule base;

[0039] Based on the execution order and logic of the rules, write and execute SQL query statements or call data processing programs to scan and analyze data in the local database;

[0040] The evaluation dimensions include, but are not limited to: data integrity (checking for null values ​​and missing values), data accuracy (validating data format, value range, and business logic), data consistency (checking for cross-table constraints and logical contradictions), data timeliness (checking for data update timestamps), and data uniqueness (checking for duplicate records).

[0041] Record the evaluation results for each rule, including the number of records checked, the number of anomalies, the anomaly ratio, the anomaly record ID, and the specific anomaly value.

[0042] As a preferred technical solution of the present invention, the structure of the data quality assessment report includes:

[0043] Report header information: Node ID, evaluation time, evaluation rule version;

[0044] Assessment Overview: Overall score, scores for each dimension, total number of anomalies;

[0045] Detailed evaluation results: List the evaluation indicators, expected values, actual values, degree of deviation, and examples of abnormalities according to the rules.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] 1. Completely resolves the conflict between privacy and quality assessment: It perfectly aligns with the "visible but unusable" principle of the trusted data space. All operations involving the original data are completed locally at the data source, fundamentally eliminating the risk of original data leakage, while accurately assessing data quality.

[0048] 2. High accuracy and adaptability: Through a rule generation mechanism that prioritizes automatic matching and supplements it with manual customization, the system can flexibly adapt to the quality assessment needs of various heterogeneous data sources and complex business scenarios, ensuring the applicability and accuracy of the assessment rules. The rule base has self-learning and growth capabilities.

[0049] 3. Global perspective and centralized control: The central console can efficiently summarize and deeply analyze the evaluation results of all nodes, providing data space operators with unprecedented global data quality visualization capabilities and control methods, helping to improve the data asset level and trustworthiness of the entire data space;

[0050] 4. High system security: From metadata upload and rule download to report feedback, the entire process adopts high-strength encryption and integrity verification mechanisms to ensure the security and trustworthiness of the assessment process itself and prevent assessment instructions or results from being stolen or tampered with.

[0051] 5. High degree of automation and good scalability: Standardized methods and processes facilitate automated deployment and execution, significantly reducing manual operation costs and the probability of errors. The system architecture design has good horizontal scalability, supporting large-scale node access. Attached Figure Description

[0052] Figure 1 A flowchart for a data quality assessment method for trusted data space nodes;

[0053] Figure 2 A flowchart for generating evaluation rules for the data quality assessment method of trusted data space nodes. Detailed Implementation

[0054] 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.

[0055] Please see Figure 1 and Figure 2 This invention provides a technical solution: a method for assessing the data quality of trusted data space nodes, comprising the following steps:

[0056] Each node extracts metadata from its database through a locally deployed data quality assessment service. The metadata includes at least one of the following: data structure information, field type, data format, constraints, data volume statistics, and update time, but does not contain any original data content.

[0057] The nodes encrypt the metadata and upload it to the central control console through a secure communication channel;

[0058] The central console receives and decrypts metadata, parses its data characteristics, and matches it against a pre-defined evaluation rule base. The matching process is achieved through multi-dimensional feature similarity calculation. If the matching similarity is lower than a pre-defined threshold, a manual rule customization process is initiated. The rule base is constructed as follows:

[0059] The data quality assessment rule base maintained by the central console server contains multiple rule subsets, each corresponding to a type of data or business scenario. Each assessment rule in the rule base includes a rule identifier, a description of applicable data characteristics, assessment dimensions, assessment metrics, predefined rule values, and assessment logic. The predefined rule values ​​are thresholds or standard values ​​preset based on historical data, industry standards, business needs, and other factors to determine whether data quality meets the standards. For example, for data integrity assessment, the predefined rule value might be set as "data field missing rate not exceeding 5%"; for data accuracy assessment, for the ID card number field, the predefined rule value might be "complies with the 18-digit ID card number encoding rules".

[0060] The central console encapsulates and encrypts the successfully matched or customized evaluation rule configuration files. The configuration files include rule identifiers, evaluation dimensions, evaluation metrics, rule predefined values, evaluation logic, execution order, and checksums, and are then distributed to the corresponding nodes through a secure channel.

[0061] After receiving and verifying the configuration file, the node performs a data quality assessment operation locally, analyzes the local data according to the assessment rules, and generates a data quality assessment report that includes assessment scores, anomaly records, and deviation analysis.

[0062] The node encrypts the assessment report and uploads it to the central control console;

[0063] The central console aggregates the evaluation reports from each node, performs comprehensive data quality analysis, generates a global data quality view, and supports visualization and early warning prompts.

[0064] In this embodiment, the metadata extraction process specifically includes:

[0065] Connect to the local database through the data quality assessment service to read metadata information from system tables or data dictionaries;

[0066] The extracted data table structure information includes table name, field name, data type, data length, whether it can be nullable, default value, and primary and foreign key relationships;

[0067] The extracted data statistics include the total number of records, the proportion of null values, the number of unique values, the range of numerical distribution, and the update frequency;

[0068] The extracted metadata is standardized and converted into a uniform JSON or XML format for easy transmission and parsing.

[0069] Metadata generation and secure upload phase:

[0070] Each node connects to the local database via a lightweight agent. This agent does not read any raw business data; instead, it extracts metadata describing the data structure and statistical characteristics by querying the database's system directory tables (such as Oracle's ALL_TAB_COLUMNS), information schemas (such as MySQL's INFORMATION_SCHEMA), or utilizing the database's metadata API. The extracted information is detailed, including: table name, field name, data type, precision, scale, nullability, default value, primary / foreign key constraints, index information, recent record count changes, estimated percentage of null fields, and approximate distribution range of null fields (such as maximum and minimum values, which can be estimated using approximation algorithms like HyperLogLog without exposing the specific values). The extracted metadata is serialized into a standard format (such as JSON Schema) and then encrypted using an asymmetric encryption algorithm (such as RSA-2048), with the public key distributed by the central console. The encrypted metadata is then transmitted to the central console via a TLS-based secure channel.

[0071] In this embodiment, the metadata encryption process uses an asymmetric encryption algorithm. Specifically, the node uses the public key provided by the central console to encrypt the metadata using either RSA or ECC. The central console uses its private key to decrypt the received encrypted metadata. To further enhance security, the metadata is compressed before encryption to reduce the amount of data transmitted.

[0072] In this embodiment, after the central console receives the ciphertext and decrypts it with its private key, it starts the metadata parsing engine. The engine extracts key feature vectors from the metadata, such as {data source type: "MySQL", table name pattern: "cust_*", fields: ["age", "income"], field "age" type: "INT", field "income" type: "DECIMAL(10,2)", ..}. Subsequently, the rule matching engine calculates the similarity between the feature vectors and the rule templates in the rule base. The multi-dimensional feature similarity calculation in the rule matching process is evaluated using the following formula:

[0073] S(M,R)=

[0074] Where S(M,R) represents the overall matching similarity between metadata M and rule R, w i sim(f) represents the weight of the i-th feature term, whose value is determined by expert experience or a machine learning model based on the importance of the feature. i M ,f i RS(M,R) is the similarity function of feature i. For numerical features (such as data length), the reciprocal of the relative error can be used. For categorical features (such as data type), the Jaccard similarity coefficient or cosine similarity can be used for calculation. Set the similarity threshold θ=0.85. When S(M,R)≥θ, it is considered a successful match. Otherwise, the match fails.

[0075] In this embodiment, the manual rule customization process includes:

[0076] The central console pushes the metadata feature information of unmatched rules to the manual rule customization platform;

[0077] The platform notifies the technical staff, who can then view the detailed feature information of the metadata.

[0078] Based on data business scenarios, industry standards, and historical experience, technical personnel create or modify evaluation rules on the platform, defining evaluation dimensions, indicators, predetermined values, and logic.

[0079] After the newly customized rules are tested and verified, they are added to the evaluation rule base in the central console and associated with the corresponding metadata features for subsequent automatic matching.

[0080] In this embodiment, the encapsulation and encryption of the evaluation rule configuration file specifically includes:

[0081] The configuration file is encapsulated using a specific structure, including a rule metadata area and a rule body area;

[0082] The rule metadata area includes the rule ID, version number, applicable data type, creation time, and author;

[0083] The rule body includes the specific evaluation condition expression, threshold parameters, and execution script;

[0084] After matching or customization is completed, the central console generates the corresponding evaluation rule configuration file. This file, defined in XML or YAML format, clearly describes the rule ID, name, applicable evaluation dimensions (such as accuracy), specific evaluation logic (such as an SQL query snippet: SELECT COUNT(*) FROM table WHERE column NOT BETWEEN min_val AND max_val), threshold parameters (min_val=0, max_val=100), and anomaly detection conditions (such as anomaly count > 0). To ensure the integrity and tamper-proof nature of the rules during transmission, the system uses the SHA-256 algorithm to generate a hash checksum for the configuration file content. Subsequently, a hybrid encryption mechanism is employed: first, a randomly generated AES-256 key is used to encrypt the configuration file itself; then, the target node's public key is used to encrypt this AES key. Finally, the encrypted configuration file and the encrypted AES key are sent to the node. Upon receiving the file, the node first decrypts the AES key using its own private key, then decrypts the configuration file using the AES key, and finally recalculates the hash value and compares it with the sent checksum. After ensuring the rule is correct, it is stored in the local rule base.

[0085] Use a hash algorithm (such as SHA-256) to generate a digital digest of the complete configuration file content as a checksum;

[0086] The central control console uses a symmetric encryption algorithm (such as AES-256) to generate a random key to encrypt the configuration file, and then uses the public key of the target node to encrypt the random key to form the final encrypted data packet.

[0087] In this embodiment, the process of a node receiving and decrypting a configuration file includes: after receiving an encrypted data packet, the node first uses its own private key to decrypt and obtain a symmetric key; it then uses the symmetric key to decrypt and obtain the plaintext of the configuration file; the node locally calculates the hash value of the configuration file content and compares it with the received checksum; if they match, the verification passes and the rules are stored in the local rule base; if they do not match, the node sends an error report to the central console and requests a retransmission.

[0088] In this embodiment, the node performs a data quality assessment operation locally, including:

[0089] The data quality assessment service on the node loads assessment rules from the local rule base;

[0090] Based on the execution order and logic of the rules, write and execute SQL query statements or call data processing programs to scan and analyze data in the local database;

[0091] The evaluation dimensions include, but are not limited to: data integrity (checking for null values ​​and missing values), data accuracy (validating data format, value range, and business logic), data consistency (checking for cross-table constraints and logical contradictions), data timeliness (checking for data update timestamps), and data uniqueness (checking for duplicate records).

[0092] Record the evaluation results for each rule, including the number of records checked, the number of anomalies, the anomaly ratio, the anomaly record ID, and the specific anomaly value.

[0093] In this embodiment, the structure of the data quality assessment report includes:

[0094] Report header information: Node ID, evaluation time, evaluation rule version;

[0095] Assessment Overview: Overall score, scores for each dimension, total number of anomalies;

[0096] Detailed evaluation results: List the evaluation indicators, expected values, actual values, degree of deviation, and examples of abnormalities according to the rules.

[0097] The following example can be used as a reference in practical applications:

[0098] I. Applications in the Trusted Data Space of the Financial Industry

[0099] Suppose that multiple banks are connected as nodes in a certain trusted financial data space, and they want to share customer tag data for joint risk control modeling, but they need to verify the data quality of each other.

[0100] Step S101: The agent program of bank node A extracts the metadata of its customer_risk_label table, including fields (such as cust_id, risk_score, update_time), type, approximate maximum / minimum value of the risk_score field (obtained through database statistics), and the proportion of null values. This is then uploaded after being encrypted using the public key from the central console.

[0101] Step S102: The central console decrypts the metadata and parses the features: Table name: "customer_risk_label", containing fields: ["risk_score"], "risk_score" type: NUMBER(3,0). The rule engine matches it against the rule base. It finds a rule R007 described as "risk score table score range check", with a feature similarity of 0.92 (>0.85), and the match is successful. The rule logic is: the risk score should be between 0 and 999.

[0102] Step S103: The central console encrypts the configuration file of rule R007 (containing the check SQL logic and thresholds) using the public key of bank node A and then distributes it.

[0103] Step S104: Bank A node decrypts, verifies, and stores the rules. The agent program executes the SQL locally: SELECT COUNT(*) FROM customer_risk_label WHERE risk_score NOT BETWEEN 0 AND 999, and finds 5 abnormal records.

[0104] Step S105: Bank A generates an assessment report, pointing out 5 anomalies in the "Accuracy" dimension, with a score of 98. The report is then encrypted and uploaded.

[0105] Step S106: The central console summarizes the reports from all bank nodes, showing that Bank A's data quality is excellent, while Bank B's similar data has a large number of null values. The space administrator can use this information to prompt Bank B to carry out data governance.

[0106] II. Rule-making Scenarios in the Healthcare Industry

[0107] In a certain medical data space, a specialized hospital has for the first time accessed a data table named gene_sequence_meta generated by a new type of gene sequencer.

[0108] Step S201: After the hospital node uploads its metadata, the central console rule engine fails to find a rule with high similarity from the existing rule base (the highest similarity is only 0.65).

[0109] Step S202: The system automatically triggers the manual rule customization process and pushes the metadata feature to bioinformatics experts on the platform.

[0110] Step S203: After analysis, experts found that the table needed to be checked for "whether the gene sequence length is within a reasonable range" and "whether the specific identifier conforms to the naming conventions". Experts created two new rules, R201 and R202, on the graphical interface, defining the checking logic and parameters.

[0111] Step S204: After testing, the new rule is stored in the rule base and associated with the metadata characteristics of that type. The configuration file is encrypted and sent to the hospital node.

[0112] Step S205: The hospital node performed a local assessment and successfully identified some non-standard data.

[0113] Step S206: Subsequently, when any other node uploads similar gene sequencing metadata, the system can automatically match and apply rules R201 and R202, realizing rule sharing and reuse.

[0114] Although embodiments of the invention have been shown and described (see the detailed description above), it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for assessing the data quality of trusted data spatial nodes, characterized in that, Includes the following steps: Each node extracts metadata from its database through a locally deployed data quality assessment service. The metadata includes at least one of the following: data structure information, field type, data format, constraints, data volume statistics, and update time, but does not contain any original data content. The node encrypts the metadata and uploads it to the central control console via a secure communication channel; The central console receives and decrypts the metadata, parses its data features, and matches it with a preset evaluation rule base. The matching process is achieved through multi-dimensional feature similarity calculation. If the matching similarity is lower than a preset threshold, the manual rule customization process is initiated. The central console encapsulates and encrypts the successfully matched or customized evaluation rule configuration file. The configuration file includes rule identifier, evaluation dimension, evaluation index, rule predefined value, evaluation logic, execution order and check code, and is then distributed to the corresponding node through a secure channel. After receiving and verifying the configuration file, the node performs a data quality assessment operation locally, analyzes the local data according to the assessment rules, and generates a data quality assessment report that includes assessment scores, abnormal records, and deviation analysis. The node encrypts the evaluation report and uploads it to the central control console; The central console aggregates the evaluation reports from each node, performs comprehensive data quality analysis, generates a global data quality view, and supports visualization and early warning prompts.

2. The data quality assessment method for trusted data spatial nodes according to claim 1, characterized in that: The metadata extraction process specifically includes: Connect to the local database through the data quality assessment service to read metadata information from system tables or data dictionaries; The extracted data table structure information includes table name, field name, data type, data length, whether it can be nullable, default value, and primary and foreign key relationships; The extracted data statistics include the total number of records, the proportion of null values, the number of unique values, the range of numerical distribution, and the update frequency; The extracted metadata is standardized and converted into a uniform JSON or XML format for easy transmission and parsing.

3. The data quality assessment method for trusted data spatial nodes according to claim 1, characterized in that: The metadata encryption process employs an asymmetric encryption algorithm, specifically: the node uses the public key provided by the central console to encrypt the metadata using either RSA or ECC; the central console uses its private key to decrypt the received encrypted metadata.

4. The data quality assessment method for a trusted data space node according to claim 1, characterized in that: The multi-dimensional feature similarity calculation in the evaluation rule matching process uses the following formula: S(M,R)= 5. Among them, S(M,R) represents the overall similarity between metadata M and rule R, w i sim(f) represents the weight of the i-th feature term, whose value is determined by expert experience or a machine learning model based on the importance of the feature. i M ,f i R S(M,R) is the similarity function of feature i. For numerical features, the reciprocal of the relative error is used, and for categorical features, the Jaccard similarity coefficient or cosine similarity is used for calculation. The similarity threshold θ = 0.85 is set. When S(M,R) ≥ θ, it is considered a successful match; otherwise, the match fails.

6. The data quality assessment method for a trusted data space node according to claim 4, characterized in that: The manual rule customization process includes: The central console pushes the metadata feature information of unmatched rules to the manual rule customization platform; The platform notifies the technical staff, who can then view the detailed feature information of the metadata. Based on data business scenarios, industry standards, and historical experience, technical personnel create or modify evaluation rules on the platform, defining evaluation dimensions, indicators, predetermined values, and logic. After the newly customized rules are tested and verified, they are added to the evaluation rule base in the central console and associated with the corresponding metadata features for subsequent automatic matching.

7. The data quality assessment method for a trusted data spatial node according to claim 1, characterized in that: The encapsulation and encryption of the evaluation rule configuration file specifically includes: The configuration file is encapsulated using a specific structure, including a rule metadata area and a rule body area; The rule metadata area includes the rule ID, version number, applicable data type, creation time, and author; The rule body includes the specific evaluation condition expression, threshold parameters, and execution script; A hash algorithm is used to generate a digital digest of the complete configuration file content as a checksum. The central control console uses a symmetric encryption algorithm to generate a random key to encrypt the configuration file, and then uses the target node's public key to encrypt the random key, forming the final encrypted data packet.

8. The data quality assessment method for a trusted data space node according to claim 6, characterized in that: The process of a node receiving and decrypting a configuration file includes: after receiving an encrypted data packet, the node first uses its own private key to decrypt and obtain a symmetric key; it then uses the symmetric key to decrypt and obtain the plaintext of the configuration file; the node locally calculates the hash value of the configuration file content and compares it with the received checksum; if they match, the verification passes and the rules are stored in the local rule base; if they do not match, an error report is sent to the central console and a retransmission is requested.

9. The data quality assessment method for a trusted data spatial node according to claim 1, characterized in that: The node performs local data quality assessment operations including: The data quality assessment service on the node loads assessment rules from the local rule base; Based on the execution order and logic of the rules, write and execute SQL query statements or call data processing programs to scan and analyze data in the local database; The evaluation dimensions include, but are not limited to: data integrity, data accuracy, data consistency, data timeliness, and data uniqueness. Record the evaluation results for each rule, including the number of records checked, the number of anomalies, the anomaly ratio, the anomaly record ID, and the specific anomaly value.

10. The data quality assessment method for a trusted data space node according to claim 8, characterized in that: The structure of the data quality assessment report includes: Report header information: Node ID, evaluation time, evaluation rule version; Assessment Overview: Overall score, scores for each dimension, total number of anomalies; Detailed evaluation results: List the evaluation indicators, expected values, actual values, degree of deviation, and examples of abnormalities according to the rules.

Images