Metadata risk assessment method and device for scientific data set and electronic equipment

Abstract

The invention relates to a metadata risk assessment method and device for a scientific data set and electronic equipment. The method comprises the steps that metadata features and similar metadata of the scientific data set are acquired; obtaining a risk assessment task tree based on the metadata features and a preset risk assessment rule; obtaining a first risk index matrix based on historical metadata of the similar metadata; obtaining a second risk index matrix based on the metadata management behavior of the similar metadata; and performing risk assessment on the scientific data set based on the risk assessment task tree and the risk index matrix to generate a risk assessment result. A comprehensive and accurate metadata risk assessment method is provided by establishing a systematic risk assessment task tree and comprehensively utilizing historical data and managed risk indexes, the metadata risk assessment method can effectively identify various metadata risks in metadata management of a scientific data set, the problem of comprehensively assessing potential risks is solved, and the risk assessment efficiency is improved. And powerful support is provided for security management of metadata.

Description

Technical Field

[0001] The present invention relates to the field of data risk assessment, and more specifically, to a metadata risk assessment method, device and electronic device for a scientific data set. Background Art

[0002] Scientific data mainly includes data generated through basic research, applied research, experimental development, etc. in the fields of natural sciences, engineering sciences, etc., as well as original data and its derivative data obtained and used in scientific research activities through observation, monitoring, investigation, inspection and testing. Scientific data sets are collections of scientific data organized and managed based on certain rules and standards. The management life cycle includes collection, submission, preservation, sharing and utilization, and the management process follows general standards and specific standards.

[0003] In scientific dataset management, metadata is used to describe various aspects of a dataset, such as its source, structure, content, and quality. However, with the growth of data size and the complexity of the management environment, metadata management faces many risks: metadata may be inaccurate or incomplete, leading to deviations in the interpretation and use of data; metadata quality may decline due to improper management or delayed updates, thus affecting the reliability and validity of data; and different metadata management behaviors and operations may trigger potential risks. These hidden risks are often difficult to identify and quantify by traditional methods, leading to challenges in metadata risk identification and management of scientific datasets. Summary of the invention

[0004] The present invention aims to overcome at least one defect (shortcoming) of the above-mentioned prior art and provide a metadata risk assessment method for scientific data sets to solve the problem that potential risks in metadata management are difficult to identify.

[0005] According to a first aspect of the present application, a metadata risk assessment method for a scientific dataset is provided, the method comprising: Get metadata features and similar metadata of scientific datasets; Obtain a risk assessment task tree based on metadata features and preset risk assessment rules; Based on the historical metadata of similar metadata, a first risk indicator matrix is ​​obtained; Based on the metadata management behavior of similar metadata, a second risk indicator matrix is ​​obtained; Conduct risk assessment on scientific data sets based on the risk assessment task tree and risk indicator matrix to generate risk assessment results.

[0006] By capturing the metadata characteristics of scientific datasets, referring to historical risk indicators and management behaviors of similar metadata, and building a task tree based on preset risk assessment rules, we can effectively identify various metadata risks in scientific dataset metadata management, comprehensively assess potential risk issues, and provide strong support for the security management of metadata.

[0007] Optionally, the step of obtaining a risk assessment task tree based on metadata features and preset risk assessment rules specifically includes: Identify risk factors and assessment dependencies in metadata features based on risk assessment rules; Determine the risk assessment level based on the mapping relationship between risk assessment rules and metadata features; Constructing an evaluation data-aware network based on evaluation dependencies; A risk assessment task tree is established based on risk factors, risk assessment levels and data perception networks; the risk assessment task tree includes multiple nodes, and each node information includes metadata features, risk factors, and assessment dependencies.

[0008] The risk assessment level is determined based on the mapping relationship between risk assessment rules and metadata features. By layering complex tasks with simple tasks, the system can allocate resources more efficiently, give priority to high-priority or more dependent tasks, and avoid unnecessary repeated calculations or waste of resources.

[0009] Optionally, the acquiring a first risk indicator matrix based on historical metadata of similar metadata specifically includes: Acquire historical metadata of similar metadata, perform correlation analysis based on metadata feature changes of the historical metadata to obtain correlation coefficients between metadata features of the historical metadata, and construct a metadata feature association matrix based on the correlation coefficients; the metadata feature association matrix includes metadata associations and metadata type associations; According to metadata feature types, the latest metadata of similar metadata are respectively obtained, and cluster analysis is performed on the latest metadata to obtain reference values ​​of metadata of different types; the metadata feature types include one or more of descriptive features, usage features, and quality features; The similarities between metadata features of different types and metadata reference values ​​are calculated, and a first risk indicator matrix is ​​generated according to the correlation coefficients and the similarities.

[0010] By analyzing historical data with similar metadata, we can understand the risks encountered by similar datasets in the past management process, and then provide a reference standard for the risk assessment of the current dataset. Data errors, inconsistencies, or other quality issues that may be contained in historical metadata can help identify potential risks that may exist in the current dataset. The historical management behavior and assessment results of similar metadata can be used to calibrate and optimize the current risk assessment rules, making the assessment more in line with actual application needs.

[0011] Optionally, the metadata management behavior based on similar metadata to obtain a second risk indicator matrix specifically includes: Obtain the metadata management behavior semantics of similar metadata, and obtain the metadata management behavior features through semantic feature clustering analysis; The metadata features associated with the metadata management behavior features are obtained, and according to the mapping relationship between the metadata features and the risk assessment rules, a second risk indicator matrix containing the metadata management behavior scores is obtained.

[0012] The metadata management behavior score is obtained based on manual annotation or behavior sentiment calculation.

[0013] Optionally, the risk assessment of the scientific data set is performed based on the risk assessment task tree and the risk indicator matrix to obtain the risk score matrix of the assessment object, specifically including: Determine the risk reference value of each node according to the characteristics of each node in the risk assessment tree and the risk indicator matrix; Calculate the risk score of each node based on the risk reference value and risk factor of each node; The risk characteristics and risk scores of each node are summarized to obtain the risk score matrix of the assessment object.

[0014] By analyzing the correlation between metadata management behavior characteristics and metadata features and constructing a correlation matrix, we can find out which metadata features are closely related to specific management behavior characteristics.

[0015] Optionally, the calculating the risk score of each node based on the risk reference value and risk factor of each node specifically includes: Based on the mapping relationship between metadata features and risk assessment rules, the corresponding risk factor weights are determined; Extracting a first subset matrix related to the metadata features from the first risk indicator matrix, and weighting the associated risk factors based on the correlation coefficients between the metadata features of the first subset matrix to obtain a first weighted risk factor; Extracting a second subset matrix related to the metadata management behavior from the second risk indicator matrix, and weighting the associated risk factors based on the metadata management behavior risk score of the second subset matrix to obtain a second weighted risk factor; Calculating a risk score for each node based on the risk reference value of each node, the first weighted risk factor, and the second weighted risk factor; Based on the evaluation dependency of each node, the associated nodes are obtained, and the risk scores of the associated nodes are calculated. The risk scores of the associated nodes are iteratively updated based on the evaluation data perception network to obtain the final risk scores of each node.

[0016] The meticulousness and dynamic adaptability of metadata risk assessment are enhanced through risk factor weight setting, multi-dimensional risk indicator matrix weighting processing, and iterative update of node risk scores based on assessment dependency and perception network. Not only can the objectivity and accuracy of risk assessment results be improved, but also the complexity and variability of risk factors in metadata management can be effectively addressed through real-time feedback and dynamic adjustment mechanisms, providing a more comprehensive, in-depth and flexible risk management strategy for metadata protection of scientific datasets.

[0017] Optionally, the metadata risk assessment method further includes: Determine the risk level and risk influencing factors of the assessment object based on risk assessment rules and the risk score matrix of the assessment object; Based on risk levels and risk influencing factors, a risk assessment report is generated using a large language model.

[0018] By combining risk assessment rules with the risk scoring matrix, we can more accurately divide risk levels and identify key risk influencing factors. We use advanced large language model technology to generate a risk assessment report, which can intuitively display the risk status of the assessment object, allowing users to provide more intuitive data support and reference for subsequent risk management and decision-making.

[0019] According to a second aspect of the present application, a metadata risk assessment device for a scientific data set is provided, the device comprising: A task tree construction module is used to construct a risk assessment task tree based on metadata features and preset risk assessment rules; A first risk indicator matrix acquisition module, used to acquire a first risk indicator matrix based on historical metadata of similar metadata; A second risk indicator matrix acquisition module, used for acquiring a second risk indicator matrix based on metadata management behavior of similar metadata; The risk assessment result generation module is used to perform risk assessment on scientific data sets based on the risk assessment task tree and risk indicator matrix and generate risk assessment results.

[0020] Optionally, the device further comprises: Real-time acquisition module, used to automatically obtain relevant information of metadata management behavior, metadata management behavior timing logic and behavior feature logic specificity; The result feedback module is used to obtain feedback from users associated with scientific data sets on risk assessment results, obtain associated metadata features and risk assessment rule features based on feedback information features, and optimize risk factor weights and risk reference values; The knowledge base construction module is used to classify risk assessment results according to metadata type and metadata feature clustering, extract semantic features of risk assessment results and corresponding assessment feedback, and build a multi-classification risk feature knowledge base.

[0021] According to a third aspect of the present application, an electronic device is provided, including: a memory for storing one or more computer programs; The processor implements the metadata risk assessment method for scientific datasets according to the first aspect when the one or more computer programs are executed by the processor.

[0022] Based on any of the above aspects, the metadata risk assessment method, device and electronic device for scientific data sets provided in the embodiments of the present application have the following effective effects: 1. By establishing a systematic risk assessment task tree and comprehensively utilizing historical data and management risk indicators, a comprehensive and accurate metadata risk assessment method is provided. This metadata risk assessment method can effectively identify various metadata risks in scientific dataset metadata management, achieve a comprehensive assessment of potential risk issues, and provide strong support for the security management of metadata.

[0023] 2. By collecting metadata management behaviors and feedback in real time, the risk assessment process can be dynamically optimized, thereby improving the accuracy and practicality of risk assessment results, helping data managers to develop more effective management strategies, and improving data management quality and decision-making levels. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 A schematic diagram of an application scenario of the metadata risk assessment method for a scientific dataset provided in this embodiment.

[0026] Figure 2 This is a flowchart of the metadata risk assessment method for a scientific dataset provided in this embodiment.

[0027] Figure 3This embodiment provides a schematic flow chart of the sub-steps of step S2.

[0028] Figure 4 This embodiment provides a schematic flow chart of the sub-steps of step S3.

[0029] Figure 5 This embodiment provides a schematic flow chart of the sub-steps of step S4.

[0030] Figure 6 This embodiment provides a schematic flow chart of the sub-steps of step S5.

[0031] Figure 7 This embodiment provides a schematic flow chart of the sub-steps of step S52.

[0032] Figure 8 This embodiment provides a flowchart of steps S61 to S62.

[0033] Fig. 9 A schematic diagram of the functional modules of the metadata risk assessment device for scientific datasets provided in this embodiment.

[0034] Fig.10 A schematic diagram of the structure of an electronic device provided by this embodiment. DETAILED DESCRIPTION

[0035] The drawings of this application are only used for illustrative purposes and should not be construed as limiting the present application. In order to better illustrate the following embodiments, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product; it is understandable to those skilled in the art that some well-known structures and their descriptions in the drawings may be omitted.

[0036] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0037] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] The existing risk assessment methods usually focus on static data and cannot fully reflect the risks of dynamic changes. Traditional methods have difficulty in dealing with the risks of unstructured data, and the assessment of complex data management behaviors is also relatively limited. This leads to challenges in metadata risk identification and management of scientific data sets.

[0039] This embodiment provides a technical solution that can solve the above-mentioned problem. The specific implementation methods of this application are described in detail below in conjunction with the accompanying drawings.

[0040] For example, a schematic diagram of an application scenario of a metadata risk assessment method for a scientific data set provided in an embodiment of the present application is provided. Figure 1 As shown, the application scenario at least includes a server 100 and a terminal 200 that can communicate with the server 100. The server 100 has an image processing function and may also have a data processing function; the terminal device 200 has a streaming media playback function and may also have a data display function.

[0041] It is understandable that the server 100 can be an independent electronic device or a cluster of multiple electronic devices; the terminal 200 can be a smart phone terminal, a personal computer, a tablet computer, a car terminal, etc., but is not limited thereto.

[0042] In one practicable manner, the server 100 and the terminal 200 may respectively execute the metadata risk assessment method for a scientific dataset provided in the embodiment of the present application, or, optionally, the metadata risk assessment method for a scientific dataset provided in the embodiment of the present application is partially executed in the server 100 and partially executed in the terminal 200.

[0043] like Figure 2 As shown, this embodiment provides a metadata risk assessment method for a scientific data set, which may include the following steps: S1. Obtain metadata features and similar metadata of scientific datasets.

[0044] In the specific implementation process, the metadata characteristics of the dataset can be classified by analyzing the existing metadata standards and management specifications of scientific datasets. Specifically, metadata characteristics include descriptive characteristics, usage characteristics, and quality characteristics.

[0045] Descriptive features include fields related to the basic information of the data itself, which are used to describe the source, content, and purpose of the data. Descriptive features include: dataset name, dataset keywords, dataset domain, data description, data generation date, etc.

[0046] Usage characteristics refer to the characteristics that reflect how datasets or metadata are used, accessed, and operated, including: access rights, usage methods, auxiliary software tool names, sharing types, access frequency, number of downloads, protection period, etc. Among them, sharing types include unconditional sharing, conditional sharing, and temporary non-sharing.

[0047] Quality characteristics are used to evaluate the accuracy, consistency and availability of metadata, including: latest publication date, update frequency, spatial scope, reference standard specifications, collection method, verification method, etc.

[0048] Exemplarily, the scientific dataset includes a genomic scientific dataset, and the descriptive features of the genomic scientific dataset include the dataset name, dataset keywords, and the field to which the dataset belongs; the usage features include application direction, sharing type, and acquisition method; and the quality features include collection method, verification method, sample source, and data consistency.

[0049] Similar metadata refers to metadata with similar characteristics to scientific datasets. The similarity between metadata can be reflected in the data type, application field, usage frequency, quality standards, etc. For example, data type similarity can be obtained based on the semantic similarity of metadata fields such as data type, data field, data source, etc.; application field similarity can be obtained based on the similarity of high-frequency access user characteristics associated with the dataset, and the similarity of derived data or derived results characteristics; usage frequency similarity can be obtained based on the similarity of short-term high-frequency access at a specific time and for specific needs.

[0050] S2. Construct a risk assessment task tree based on metadata features and preset risk assessment rules.

[0051] During the specific implementation process, preset risk assessment rules can be established based on standard specifications and management experience, including descriptive risk rules, usage risk rules and quality risk rules.

[0052] Descriptive risk rules can be derived from general management specifications or basic management requirements, and can be established based on standard specifications such as data collection and production, data classification, data description basis, data naming, etc. For example, the source of data should be traceable, the classification of the field to which the data belongs should have a reference basis, and the data description should fully reflect all the data content of the data set.

[0053] Risk usage rules can be established based on user access control policies, data classification standards, data interaction behavior monitoring statistics, and metadata update policies. For example, the scientific data policy stipulates that scientific data centers should regularly evaluate the sharing of scientific data sets, and if they meet the conditions for unconditional or conditional sharing, they should promptly convert the sharing type; the data should also be promptly converted to a sharing type after being desensitized and declassified. According to the above policy provisions, in actual business, specific regular assessments can include assessments of data integrity and update frequency.

[0054] Quality risk rules can be established based on quality control measures, quality evaluation standards, data verification mechanisms, etc. For example, a data quality control measure stipulates that metadata fields such as data usage and data application direction of a data set should be updated every quarter based on the data application sharing and use records. The corresponding quality risk rule is to integrate and analyze all data application sharing and use records of the previous quarter, analyze the differences with the current subject classification, subject classification, data usage and other metadata fields, score metadata risks and make metadata management risk recommendations.

[0055] For example, in the metadata management of genomic scientific datasets, if the descriptive information such as species name or genome length of the genomic dataset is inconsistent with similar datasets, or does not conform to the standardized description format, there is a descriptive risk. If the characteristics of the target user group of the dataset vary greatly, or the application field of the dataset shifts, there is a usage risk, and it is necessary to evaluate whether it will affect the applicability of the metadata. If the sequencing depth or annotation accuracy of the genomic data is lower than the historical average or industry standard, there may be a quality risk.

[0056] like Figure 3 As shown, in the embodiment of the present invention, a risk assessment task tree is constructed based on metadata features and preset risk assessment rules, specifically including steps S21 to S24: S21. Identify risk factors and assessment dependencies contained in metadata features based on risk assessment rules.

[0057] The identification of risk factors can be carried out according to the rule types defined in the preset risk assessment rules and in combination with the risk assessment requirements at specific times and stages. Risk rules can be divided into general risk categories and specific risk categories. Among them, general risk categories specifically include data security, data quality, data consistency, data compliance, etc. Specific risk categories are customized according to specific data management needs. For example, in medical data sets, data privacy and patient information protection are key risk types, while in meteorological data sets, the timeliness and accuracy of data may be more important risk types. For different stages of scientific data set management or special stages of shared use, the focus of risk is also different. For example, in the data collection stage, more attention can be paid to the integrity and accuracy of the data; in the data sharing stage, privacy and security can become the primary risks; when responding to emergencies, timeliness and integrity may be the primary risks. General risk factor identification can be based on policies, systems or general standards and specifications. For example, the integrity of data description (data quality) can be evaluated by checking whether the metadata fields meet the description specifications; data access control (data security) can be evaluated by checking whether the access permission configuration meets the predetermined security standards. The identification of specific risk factors can be based on domain-related management requirements or dedicated institutional norms. For example, the privacy protection measures for data collection (data security) can be evaluated by checking whether there are informed consent forms and data desensitization records from participants; the timeliness of the data (data quality) can be evaluated by checking metadata update records and the latest progress in data application.

[0058] Specifically, metadata includes structured metadata and unstructured metadata, and risk factors can be extracted and identified from structured or unstructured metadata.

[0059] Structured metadata usually exists in the form of fields and tags, which is convenient for direct analysis. For example, the metadata of scientific data sets contains fields such as "data update frequency", "data source", and "access rights". For the "data update frequency" field, if the preset risk assessment rule stipulates that the data must be updated within a specific time interval, if the actual update frequency is lower than the specific time interval, the system will identify the "update delay" risk factor. Similarly, when the "data source" field shows that the data comes from an unverified source, the system can identify the "unreliable data source" risk factor, and then issue a warning about the credibility of the data.

[0060] For example, in the management of genomic data sets, the preset risk rules stipulate that genomic data must be updated within a certain period of time according to the latest sequencing technology, and the actual update frequency field and related metadata update records do not meet the requirements, then the "update delay" risk factor will be identified, and the "update delay" risk factor may cause the data results to be outdated or inaccurate.

[0061] For the identification of risk factors of unstructured metadata features, semantic analysis can be used to parse unstructured content such as data description documents, submission plans, quality reports, and application cases to identify the corresponding risk factors. For example, if the data description in the quality report is inconsistent with the actual data, the system will identify the risk factor of "data inconsistency"; if the storage format or transmission method of the data does not match the storage format or transmission method specified in the submission plan, there will be risk factors of "non-compliant storage format" or "transmission security risks"; similarly, in some application cases, if the data is used improperly or the results are not as expected, the system will identify the risk factors of "misuse of data" or "unreliable results". Identifying these risk factors helps ensure the standardization and reliability of data management and use.

[0062] Taking genomic data management as an example, data quality reports usually contain information such as sequencing accuracy, coverage, and variant detection results. If the system finds that the description in the quality report does not match the actual sequencing data, the system will identify the risk factor of "data inconsistency". Based on this risk factor, the system can prompt that the accuracy of the data needs to be further verified.

[0063] Assessment dependencies refer to the interrelationships and dependencies between different metadata features or risk factors. These dependencies determine that when evaluating a certain metadata feature or risk factor, the impact of other related metadata features or risk factors must be considered. For example, when evaluating the data integrity of metadata, it is necessary to consider the accuracy of both the data collection method and the data processing process. If there are irregular operations in the data processing process, even if there is no problem with the data collection method, it may cause risks to data integrity. Therefore, when constructing a risk assessment task tree, these dependencies need to be considered to ensure the comprehensiveness and accuracy of the risk assessment.

[0064] Taking genomic data management as an example, genomic data involves personal privacy. In the risk assessment process, the "data privacy" risk factor may depend on the setting of "access rights". If the metadata characteristics show that the access rights are not properly managed (for example, the rights are too broad or not updated in time), it may lead to the risk of privacy leakage. Therefore, when constructing a risk assessment framework for genomic data, these dependencies must be fully considered to ensure that when assessing data privacy risks, the rationality and effectiveness of access rights can be comprehensively and accurately assessed, thereby ensuring the security and compliance of data management.

[0065] S22. Determine the risk assessment level based on the mapping relationship between the risk assessment rules and the metadata features; The mapping relationship between risk assessment rules and metadata features refers to associating each metadata feature with the corresponding risk factor and assessment rule. For example, the "data format" metadata feature is mapped to the "format non-compliance" risk factor, and the "data update frequency" metadata feature is mapped to the "data obsolescence" risk factor.

[0066] Specifically, there is a one-to-one correspondence between metadata feature classification and metadata management risk rule classification: descriptive features correspond to descriptive risk rules, usage features correspond to usage risk rules, and quality features correspond to quality risk rules.

[0067] In the specific implementation process, the mapping relationship between metadata features and risk rule features can be established according to the evaluation and analysis object of the rule, and then the corresponding risk weights can be assigned to the metadata features according to the number of metadata feature types contained in each rule and the number of rule feature types corresponding to each metadata feature. According to the importance of the metadata feature and the risk level of the mapping rule, the evaluation priority of each feature can be determined.

[0068] When determining the evaluation level, first, all evaluation tasks are divided into different categories and priorities according to the metadata features and the corresponding risk factors. Secondly, the dependencies between the tasks are analyzed, and the evaluation tasks with more or stronger dependencies are divided into higher levels. For example, the evaluation of "data processing standardization" may depend on the evaluation results of "data collection methods". If there are risks in the data collection method, the evaluation of data processing standardization needs to be adjusted accordingly. Finally, the execution efficiency of the evaluation task is also an important factor in determining the evaluation level. The evaluation of some metadata features may be more complex or require more computing resources than other features. These evaluation tasks should be placed at a higher level. Simple, directly related evaluation tasks can be arranged for evaluation at a lower level.

[0069] Through such hierarchical division, the system can allocate resources more efficiently, give priority to high-priority or highly dependent tasks, and avoid unnecessary repeated calculations and resource waste. This not only improves the overall efficiency of the evaluation work, but also ensures the accuracy and reliability of the evaluation results.

[0070] S23, constructing an evaluation data perception network based on evaluation dependency; The purpose of building an assessment data perception network is to establish a network system that dynamically perceives and processes risk assessment data by identifying and utilizing the assessment dependencies between metadata features. The assessment data perception network enables each assessment task to collaborate with each other during risk assessment and perceive the changes in data and the evolution of risk factors in real time, thereby improving the accuracy and efficiency of risk assessment. First of all, the cornerstone of the construction process is that the system must comprehensively and accurately identify the dependencies between all assessment tasks. These dependencies constitute the logical basis of the perception network and determine the direction of information flow and the triggering order of assessment tasks.

[0071] The basis for building a perception network is to identify the dependencies between evaluation tasks. According to the identified evaluation dependencies, multiple data perception nodes are set up in the perception network. These data perception nodes are the trigger points and information collection points of the evaluation tasks. Each node focuses on monitoring the status of specific metadata features. Once any anomalies or changes are detected, the corresponding evaluation tasks are immediately triggered, and the results are quickly passed to all nodes in the network that rely on the results. In this way, these data perception nodes form an interconnected perception network through data flow and information sharing.

[0072] The evaluation data perception network is not only a static architecture, but also a system that can dynamically perceive data changes. The evaluation data perception network can capture the changes of all metadata features in the data set in real time. Whether it is data update, deletion or modification, the data perception node can capture and adjust the relevant evaluation tasks in real time. For example, when the "data update frequency" data perception node detects that the data has not been updated for a long time, it will immediately start the corresponding "data obsolescence" risk assessment task, and feed back the assessment results to the "data integrity" assessment node in real time to correct the assessment basis of the "data integrity" assessment node. This dynamic perception mechanism enables the system to respond to data changes in a timely manner and adjust the corresponding risk assessment.

[0073] S24. Establish a risk assessment task tree based on metadata features, risk factors, assessment dependencies, risk assessment levels and data perception networks; the risk assessment task tree includes multiple nodes, and each node information includes metadata features, risk factors, and assessment dependencies.

[0074] Node priorities can be determined based on risk impact, user needs and feedback, data sensitivity and confidentiality, data quality and update, and the complexity of technical implementation. By comprehensively considering these factors, the risk management process can be effectively optimized to ensure the overall quality and security of scientific data sets.

[0075] Specifically, according to the degree of risk impact, the node priority can be divided into high, medium, and low impact risk nodes. If problems with the node will directly cause serious impact on the accuracy, compliance or availability of the data, then these nodes are classified as high impact risk nodes, and these nodes are correspondingly high priority. For example, data consistency errors will directly affect the credibility of research conclusions, so such nodes are high impact risk nodes. If the problem of the node has a large impact on the data, but it can be alleviated or corrected by other means, it is considered a medium priority. For example, improper access permission settings may lead to data security risks, but can be solved through permission review and update, then this node is a medium impact risk node. If the problem of the node does not directly affect the scientificity or usability of the data, it is classified as a low priority. For example, insufficiently detailed data descriptions may affect the user experience, but will not directly affect the core quality of the data, then this node is a low impact risk node.

[0076] In this embodiment, the node priority may also be determined based on user needs and feedback.

[0077] Determine node priority based on user usage frequency: For datasets or features that are frequently accessed by users, their priority can be classified as higher, because these data are of great significance in the user's research or application. On the contrary, if the user accesses a dataset or feature less frequently, its priority can be appropriately reduced.

[0078] Determine node priority based on user feedback: Problems and suggestions from users during use will affect the priority of nodes. For example, if users report that a feature is used less frequently, this may indicate that the relevance of the feature needs to be re-evaluated, and its priority may need to be lowered.

[0079] Node priority can also be determined based on the sensitivity and confidentiality of the data. For example, the management of sensitive data such as personal privacy and commercial secrets requires a higher priority to ensure data security and legitimacy.

[0080] Node priority can also be determined based on data quality and update status. Frequently updated data sets require stricter quality control and will have a higher priority to ensure data consistency during the update process. Nodes with more problems in historical data will also have a higher priority, as these problems may affect future data processing and analysis.

[0081] Node priority can also be determined based on the complexity of technical implementation. By evaluating the complexity of technical implementation, potential risk nodes can be identified more effectively. For example, metadata inconsistencies and quality issues caused by improper system integration or lack of effective standardization and verification processes. The complexity of technical implementation directly affects the construction of the risk assessment task tree and the prioritization of nodes. Focus should be placed on nodes with greater technical difficulty and higher potential risks to reduce metadata risks of scientific datasets.

[0082] Taking the construction of metadata risk assessment task tree of corn variety approval scientific data set as an example, feature nodes include description feature node, usage feature node and quality feature node. Specifically, description feature node includes dataset name, data description and dataset keywords. Usage feature node includes access rights, sharing type and usage frequency. Quality feature node includes data consistency, update frequency and verification method.

[0083] S3. Obtain a first risk indicator matrix based on historical metadata of similar metadata.

[0084] Historical metadata with similar metadata refers to metadata with similar characteristics to the current scientific dataset, and these metadata come from the historical records of scientific dataset management. The similarity of metadata can be reflected in data type, application field, usage frequency, quality standards, etc. Specifically, data type similarity can be obtained based on the semantic similarity of metadata fields such as data type, data field, and data source. Data application field similarity is obtained based on the similarity of high-frequency access user characteristics associated with the dataset, and the similarity of derived data or derived results characteristics. The usage frequency includes the similarity of short-term high-frequency access at a specific time and for specific needs. For example, if the environmental monitoring dataset, viral genome dataset, and Chinese medicinal material dataset are frequently accessed by the same or similar users at the same time in a certain period of time, then the fields such as data usage and data sharing methods of the relevant datasets may need to be modified. In the process of metadata management and risk assessment, these similar historical metadata can be used to predict and judge the potential risks, applicability, and sharing possibilities of the current metadata.

[0085] By analyzing past management experience, usage feedback, and risk events of similar metadata, we can more accurately assess whether current metadata is suitable for unconditional sharing or requires further protection measures.

[0086] For example, in metadata management of genomic scientific datasets, if the dataset to be evaluated is a genome sequence of a certain species, then historical metadata related to genome datasets of the same species is considered historical metadata of similar metadata.

[0087] By analyzing historical data with similar metadata, we can understand the risks encountered by similar datasets in the past management process, and thus provide a reference standard for the risk assessment of the current dataset. Historical metadata may contain data errors, inconsistencies, or other quality issues. By referring to the quality issues contained in historical metadata, we can help identify potential risks that may exist in the current dataset.

[0088] like Figure 4 As shown, step S3 specifically also includes S31 to S33.

[0089] S31. Acquire historical metadata of similar metadata, perform correlation analysis based on metadata feature changes of the historical metadata to obtain correlation coefficients between metadata features of the historical metadata, and construct a metadata feature association matrix based on the correlation coefficients; the metadata feature association matrix includes metadata associations and metadata type associations.

[0090] In this implementation, correlation analysis is performed based on the metadata feature changes of the historical metadata, and the correlation coefficients between the metadata features are calculated. In the specific implementation process, the correlation coefficients between the metadata features can be calculated using the Pearson correlation coefficient or other statistical methods.

[0091] Metadata association refers to the association between metadata field contents of the same feature type. The association relationship of metadata of the same feature type can be obtained from a single or adjacent data management life cycle or the data interaction behavior of a single role user. For example, the association between the quality feature "data update frequency" and the quality feature "data collection cycle" can be obtained from the data quality report and data submission plan submitted by the data producer.

[0092] Metadata types include descriptive metadata, technical metadata, quality metadata, operational metadata, and management metadata. Metadata type association refers to the association between descriptive metadata, technical metadata, quality metadata, operational metadata, and management metadata. For example, the descriptive metadata "data set target field: clinical diagnosis and treatment" describes the target field of the data, the technical metadata "data destruction policy: data retention for 3 years, destroyed according to XX standards" defines the technical specifications for data processing, and the management metadata "access control: internal use, encrypted access" stipulates the security and access rights of the data. By constructing an association matrix between metadata types, the dependencies between different types of metadata and the potential risk propagation paths between different types of metadata can be identified. For example, errors in technical metadata may affect the accuracy of management metadata, and the association between technical metadata and descriptive metadata may affect the reproducibility and verifiability of the data.

[0093] Through metadata type association, metadata type associations other than metadata feature associations can be identified, which provides a clearer identification of cross-type risk transmission. For example, by analyzing the association between descriptive metadata and technical metadata, if it is found that the target field of a dataset is clinical diagnosis and treatment, the dataset needs to comply with specific laws and regulations, such as data protection and destruction standards. This allows the system to prioritize the inspection and collection of data that meets these requirements when acquiring data, thereby improving the accuracy and compliance of data acquisition.

[0094] S32. Acquire the latest metadata of similar metadata according to metadata feature types, and obtain reference values ​​of metadata of different types through metadata feature clustering analysis; the metadata feature types include one or more of descriptive features, usage features, and quality features.

[0095] Metadata characteristics can include descriptive characteristics, usage characteristics, and quality characteristics.

[0096] Different types of metadata features are analyzed through clustering algorithms (such as K-means, hierarchical clustering, or DBSCAN, etc.), and the latest metadata is divided into different clusters according to the metadata features. After the data features in each cluster are analyzed, the system will calculate the center point or average value of the cluster as the reference value of the cluster metadata. These reference values ​​will be used as benchmarks for comparison and analysis in subsequent risk assessments. Taking the meteorological data set as an example, cluster analysis found that the accuracy and consistency of temperature measurement data of a certain set of historical data is the highest when it is collected four times a day. Therefore, the frequency of collecting four times a day can be used as a reference value for this type of temperature data. In future risk assessments, this reference value will be used to compare the collection frequency of current data to determine whether there is a risk of abnormal collection frequency.

[0097] S33. Calculate the similarity between metadata features of different types and metadata reference values, and generate a first risk indicator matrix according to metadata feature correlation coefficients and similarities.

[0098] According to the reference value comparison items of each metadata type, the assessment dependencies are classified and the corresponding assessment data are extracted. For example, the similarity between the storage encryption method (operation metadata), data collection frequency (quality metadata), data access rights (management metadata) and the reference value of the current data set can be calculated to assess the risk level of the current metadata. In the specific implementation process, the similarity between the metadata feature and the metadata reference value can be calculated by methods such as Euclidean distance, cosine similarity or Manhattan distance to quantify the deviation between the current metadata and the historical reference value. The first risk indicator matrix generated by the correlation coefficient and similarity of the metadata features provides a comprehensive risk score for each metadata feature, covering the correlation with other features and its similarity with historical data, showing the risk level of each metadata feature.

[0099] S4. Obtain a second risk indicator matrix based on metadata management behavior of similar metadata.

[0100] Metadata management behaviors of similar metadata refer to the management operations and behavior records of datasets with similar characteristics to the scientific dataset to be evaluated during their life cycle. These behaviors may include the creation, modification, update, sharing, access control, data quality review, application of metadata standards, version management, backup and recovery, etc. The management behaviors of similar metadata are highly correlated with the risks that may be faced in the current metadata management process.

[0101] For example, in the metadata management of genomic scientific datasets, with the development of sequencing technology and data analysis tools, genomic scientific datasets are usually reprocessed and updated, especially in terms of data verification and quality control. Similar modification and update behaviors may involve resequencing of data, re-examination of data quality, and re-annotation of metadata. For example, if the recently collected genomic scientific datasets all use next-generation sequencing technology, adding new high-frequency common fields such as NGS sequencing platform, sequencing depth, error federation, and analysis tools, then for similar scientific datasets, it may be required to update or modify the relevant metadata fields and require sampling new technologies for resequencing. For example, after the release of new research results, the "Disease Resistance Genomic Dataset" updated the data usage field, added the newly discovered disease resistance gene information to the metadata, and redefined the application scope of the dataset. This has a strong similarity with the similar updates made to the current dataset to be evaluated after technological advances. When the analysis mentions that the current dataset has similar data application reports or cases, it can prompt the timeliness risk of the data usage field.

[0102] like Figure 5 As shown, step S4 specifically also includes S41 to S42.

[0103] S41. Obtain metadata management behavior semantics of similar metadata, and obtain metadata management behavior features through semantic feature clustering analysis.

[0104] Metadata management behavior semantics refers to the semantic information features extracted by analyzing metadata management behavior records and technical documents related to management behaviors through natural language processing technology. For example, in clinical medical data management, common behavior semantic features such as "data encryption requirements" and "update frequency requirements" can be extracted. Cluster analysis of similar management behavior information can be performed to obtain behavioral patterns, related metadata features, behavior frequencies and other behavioral features corresponding to different management behaviors. For example, the "data sharing approval" behavior may be closely related to the "data sharing method" metadata feature. By analyzing the correlation between metadata management behavior features and metadata features and constructing a correlation matrix, it can be found out which metadata features are closely related to specific management behavior features.

[0105] S42: Obtain metadata features associated with metadata management behavior features, and obtain a second risk indicator matrix containing metadata management behavior scores based on a mapping relationship between metadata features and risk assessment rules.

[0106] Metadata features associated with metadata management behavior features refer to metadata attributes or fields that are closely related to certain specific management behaviors. For example, in the data management process, the "data sharing approval" management behavior is closely related to metadata features such as "data sharing method" or "access rights". These metadata features are usually the parts that will be affected or determined by the behavior.

[0107] Specifically, the ways of obtaining metadata features associated with metadata management behavior features include feature keyword extraction, metadata field query, and preset behavior type association rule matching.

[0108] Feature keyword extraction refers to extracting core keywords or phrases from the description or record of metadata management behavior, and then identifying metadata features closely related to the behavior. For example, when a "metadata field revision" behavior is described or recorded in the system, the system will analyze this description text and extract keywords such as "field naming", "revision", "data type", etc. Through these keywords, metadata features related to this behavior can be identified, such as data field name, data type, etc. These features are metadata fields directly affected by the naming revision behavior.

[0109] Metadata field query means that in certain metadata management behaviors, the system needs to query existing metadata fields to find metadata features related to specific behaviors. For example, when performing a metadata version update, it is necessary to query the current dataset version information and update time. This type of query can directly locate metadata fields related to the version update operation in the metadata management tool or database, such as "version number", "update time", etc. These fields must be updated synchronously in the version update operation, so they are closely related to this behavior.

[0110] Preset behavior type association rule matching means that the system determines the metadata features related to specific management behaviors by matching these rules. These rules are usually pre-set based on experience or business logic to help the system automatically identify associated metadata fields. For example, when applying to change the "sharing type", the preset rules stipulate that when this behavior occurs, the metadata features related to data access rights and sharing methods need to be adjusted. By matching these preset rules, the system can automatically identify and update the corresponding metadata fields such as "access rights" and "sharing type". The rules should specifically include the rule triggering conditions and the associated behavior types, the associated metadata management behavior types and operation fields after the rule is triggered. Triggering conditions and associated behavior types, such as "When a user submits a 'data sharing type change' request, a new sharing type or data classification or classification suggestion should be identified. The associated metadata management behavior types and operation fields after the rule is triggered, such as the associated "access rights" field should be updated to the new permission level, for example, the field is changed from "XX research institution only" to "public".

[0111] In this embodiment, the metadata management behavior score is obtained based on manual annotation or behavior sentiment calculation.

[0112] For management behavior scores, each metadata feature is mapped to a specific risk indicator based on the preset risk assessment rules. For example, irregular "data sharing approval" behavior may be mapped to "data leakage risk". These scores can come from manual annotation or automatically generated through behavioral sentiment calculation. The higher the score, the greater the risk that the behavior may have.

[0113] S5. Perform risk assessment on the scientific data set based on the risk assessment task tree, the first risk indicator matrix and the second risk indicator matrix to generate risk assessment results.

[0114] like Figure 6 As shown, step S5 specifically also includes S51 to S53.

[0115] S51. Determine the risk reference value of each node according to the characteristics of each node of the risk assessment tree, the first risk indicator matrix and the second risk indicator matrix.

[0116] S52. Calculate the risk score of each node based on the risk reference value and risk factor of each node.

[0117] like Figure 7 As shown, step S52 specifically also includes S521 to S525.

[0118] S521. Determine corresponding risk factor weights based on the mapping relationship between metadata features and risk assessment rules.

[0119] The risk factor weight reflects the importance of a specific risk factor in the overall risk assessment. By analyzing the mapping relationship between metadata features and preset risk assessment rules, the corresponding risk factor weight is determined for each metadata feature. For example, in a clinical medicine dataset, the data privacy and security risk factors can have a higher weight, while the data update frequency risk factor can have a lower weight.

[0120] Specifically, determining the risk factor weights requires identifying all metadata features in the data set, and then determining the risk attributes of each metadata feature through expert consultation, industry standards, and previous risk assessment cases. Based on the metadata features, risk-related assessment rules are preset.

[0121] It is understandable that the rules for setting weights can be formulated by the data management team in combination with regulations, industry best practices, and internal organizational policies. In the specific implementation process, expert scoring methods (such as the Delphi method) or multi-criteria decision analysis methods (such as the AHP method) can be used to invite field experts to score the relative importance of each risk factor to determine the weight of each risk factor. For example, when conducting a risk assessment on a clinical medical data set, experts may determine that the weight of the data privacy and security risk factor is 0.4, while the weight of the data update frequency risk factor is 0.1.

[0122] In this embodiment, all risk factors may be standardized so that each risk factor is assigned a weight, such that the sum of the weights of all risk factors is equal to 1.

[0123] Apply weights to the risk assessment model and verify the effectiveness of the weights through historical data. The historical data includes risk events that have occurred, related metadata features, and the impact of these risk events. The historical data needs to cover different metadata feature values ​​and their corresponding risk event occurrences.

[0124] In this embodiment, statistical methods such as regression analysis can also be used to analyze the relationship between each risk factor and the actual risk event to adjust and optimize the weight. For example, if the analysis shows that the higher the weight of "data source reliability risk", the greater the probability of the actual risk event, which indicates that the importance of this factor is high. By analyzing the difference between the model prediction and the actual result (i.e., the residual), the accuracy of the model can be determined. If the difference is large, the weight of certain risk factors needs to be adjusted.

[0125] S522: Obtain a metadata feature-associated risk indicator matrix through the first risk indicator matrix, and weight the associated risk factors based on metadata feature correlation coefficients.

[0126] The first risk indicator matrix contains the risk indicator values ​​of each metadata feature, which are obtained based on similarity calculation and feature correlation analysis. By using the correlation coefficient of the metadata feature to perform weighted calculation on the associated risk factors, a more accurate feature risk indicator matrix can be obtained.

[0127] Obtain the risk indicator matrix related to metadata features, which contains the risk indicator values ​​of each metadata feature, which are obtained based on similarity calculation and feature correlation analysis. Subsequently, the system uses the correlation coefficient of the metadata features to perform weighted calculation on the associated risk factors, thereby obtaining a more accurate feature risk indicator matrix.

[0128] From the first risk indicator matrix, a subset matrix directly related to the scientific dataset metadata feature can be extracted, which contains other metadata features that are highly correlated with the scientific dataset features and their risk scores. For example, if the scientific dataset is the "data update frequency" feature in meteorological data, then features that are highly correlated with this feature (such as "data collection cycle") and their comprehensive risk scores can be extracted.

[0129] In the extracted correlation matrix, each feature will have a comprehensive risk score and a correlation coefficient with other features. The correlation coefficient is obtained by analyzing the correlation of metadata feature changes in the historical metadata dataset, reflecting how the change of one metadata feature is related to the change of another feature. When the metadata feature correlation coefficient is positive, it means that the values ​​of the two features usually increase or decrease at the same time; when it is negative, it means that when one feature increases, the other feature usually decreases. The closer the absolute value is to 1, the stronger the correlation. Features with high correlation are associated in the risk propagation process.

[0130] Categorize metadata into different types such as descriptive metadata, technical metadata, quality metadata, etc., and identify the associations between these types. For example, the descriptive metadata "data set target domain" may be associated with the quality metadata "data collection cycle", so an abnormality in one feature may lead to an increased risk of another feature.

[0131] In the metadata feature-associated risk indicator matrix, each metadata feature is associated with one or more risk factors. To weight these risk factors, the correlation between metadata features and their relationship with risk factors should be considered. Using these correlation coefficients, risk factors associated with scientific dataset metadata features can be identified. These risk factors represent the potential impact or risk propagation path between features. For each key metadata feature, its associated risk factors are identified. For example, "data collection period" may be associated with "data inconsistency risk" or "data loss risk", etc. At the same time, it is also necessary to select risk factors with strong correlation with it based on the metadata feature correlation coefficient. For example, if a metadata feature (such as "data update frequency") has a high correlation coefficient with multiple other features (such as "data collection period"), the risk factors of these related features need to be considered. Using the metadata feature correlation coefficient calculated in the previous step, the associated risk factors are weighted. The weighting process uses the correlation between the risk factor and the feature to be evaluated to amplify or reduce the weight of a specific risk factor in the overall evaluation. For example, if the correlation coefficient between the "data collection period" and "data update frequency" features is 0.8, and their risk scores are high, then these two features will be considered as key risk factors. Based on the extracted risk factors, their comprehensive risk scores are multiplied by the correlation coefficients for weighted calculation. The weighting process uses the correlation between the risk factors and the features to be evaluated to amplify or reduce the weight of specific risk factors in the overall assessment. For example, if the risk score of the "data collection cycle" feature is 0.7 and the correlation coefficient with the "data update frequency" feature is 0.8, the weighted score is 0.56. This value will be used in subsequent steps to determine the influence of this feature on the overall risk assessment.

[0132] S523: Obtain a metadata management behavior risk indicator matrix through the second risk indicator matrix, and weight the associated risk factors based on the metadata management behavior scores.

[0133] From the second risk indicator matrix, a subset matrix directly related to the metadata management behavior of scientific datasets can be extracted. This matrix contains management behavior features and their corresponding risk scores. In the extracted management behavior risk indicator matrix, the system identifies the risk factors associated with each management behavior feature based on its score. These risk factors may represent potential risk points in the metadata management process, such as "data leakage risk" or "data loss risk". According to the risk score of the extracted management behavior feature, the associated risk factors are weighted. The weighting process is based on the high and low scores of management behaviors and the correlation between these behaviors and risk factors.

[0134] In the specific implementation process, each risk factor can be assigned an initial weight based on historical data analysis and domain expert knowledge. The risk factor weight matrix composed of these weights can contain different dimensions, such as the severity of the risk factor, the frequency of occurrence, and the relevance to metadata management behavior. The risk factor weight matrix is ​​expressed as: ,in, Indicates specific metadata management behavior characteristics, Indicates the risk factor associated with it.

[0135] Select rows or columns directly related to the metadata management behavior characteristics to be evaluated from the second risk indicator matrix. These rows or columns contain risk factors and their scores directly related to these metadata characteristics. For example, for the metadata characteristic of "data backup frequency", the relevant risk factors include "backup failure risk", "data integrity risk", etc. By matching these characteristics with the risk factors in the matrix, we can find out which risk factors are related to specific metadata management behaviors. For example, feature A (data access frequency) matches risk factor 1 (data leakage risk). The higher the data access frequency, the greater the possibility of leakage, so there is a strong correlation between them. Feature B (metadata update frequency) matches risk factor 2 (data loss risk) and risk factor 3 (data consistency risk). Frequent updates may increase the risk of data loss and may also cause data inconsistency. Feature C (user permission change record) matches risk factor 1 (data leakage risk). Frequent user permission changes may increase the risk of unauthorized access, thereby increasing the possibility of data leakage. Integrate these extracted features and their corresponding risk factor scores into a new management behavior risk indicator matrix to generate a new management behavior risk indicator matrix. According to the management behavior risk indicator matrix and the weight matrix , calculate the weighted score of each risk factor as: ),in is the weight coefficient of risk factor j.

[0136] For the nodes of the scientific dataset risk assessment tree, determine their related metadata features and historical behaviors. Match the scientific dataset metadata features associated with the risk assessment tree nodes with the metadata features in the first risk indicator matrix item by item. Use vector similarity to calculate the similarity between the node features and the corresponding features in the first risk indicator matrix. Based on the matching and similarity, extract the reference value of each node in the first risk indicator matrix. Similar to the first risk indicator matrix, match the management behavior features associated with the node with the management behavior features in the second risk indicator matrix. Based on the mapping relationship between the metadata features and the management behavior features, obtain the metadata features corresponding to the scientific dataset management behavior features, thereby obtaining the metadata features potentially associated with the scientific dataset management behavior as the second risk reference value. Combine the reference values ​​extracted from the first and second risk indicator matrices to form a comprehensive risk reference value, the formula is as follows: Comprehensive risk reference value = α × first indicator reference value + β × second indicator reference value Among them, α and β are weight coefficients set according to actual conditions. The setting of weight coefficients can be based on the actual needs of metadata management of scientific data sets, including evaluating the impact of the first risk indicator matrix and the second risk indicator matrix in actual risk events based on the analysis of historical data and expert opinions; weights can also be allocated based on the criticality of metadata features (such as data type, update frequency, etc.) and the importance of management behavior features (such as access frequency, error records, etc.).

[0137] S524. Obtain the risk score of each node based on the risk reference value of each node and the weighted risk factor.

[0138] In this embodiment, the initial risk score of each node may be calculated according to the risk reference value of each node and the weighted risk factor.

[0139] Specifically, each node corresponds to several risk indicators, and the risk score of each node can be calculated using the weighted risk factor. , the calculation formula is as follows: + .

[0140] in Is a node i The risk reference value is used as the risk level benchmark value corresponding to the node; Is a node i In the j The scores on the risk indicators are It is j The risk factor weight corresponding to each risk indicator. It can be calculated based on the deviation of the scientific dataset from the reference value. For example, when evaluating the accuracy of the descriptive metadata feature "dataset name" of a scientific dataset, the reference value of similar datasets obtained from historical data is 8 (the specific score can be calculated based on the semantic similarity between the dataset name and the key paragraphs at a specific position in the supporting information carrier). The "dataset title" score of this dataset is 6. If the focus is on evaluating the difference with the specific metadata features of the subdivided field, the deviation can be calculated by subtraction as -2. If the difference with the common metadata features of all fields is to be evaluated, the score can be calculated using the distance from the score of the similar dataset title.

[0141] S525. Obtain the associated node characteristics and associated node analysis scores based on the evaluation dependencies of each node, iteratively update the associated node risk scores based on the evaluation data perception network, and obtain the final risk scores of each node.

[0142] Based on the evaluation dependency between nodes, the associated node features and associated node analysis scores are obtained to build an evaluation data perception network. The perception network will dynamically capture the risk dependency relationship between nodes and iteratively update the risk score of each node through information transmission and feedback mechanisms. This process reflects the mutual influence between nodes. For example, if the risk score of a node increases due to the high risk of its associated nodes, the system will automatically update the score of the node until the entire network reaches a stable state.

[0143] In the specific implementation process, during a special time period or emergency, while the metadata risk assessment task of a scientific dataset is being executed, the metadata management behavior of similar datasets may increase significantly. Alternatively, according to the rule triggering mechanism, a node of the risk assessment task tree is associated with another node. For example, the system detects that node 1 (data update frequency) has not been updated in the past three months, marks it as abnormal, and triggers the risk assessment task associated with node 2 (data integrity). At this time, it is necessary to analyze the characteristics of the associated nodes and calculate new analysis scores based on the state changes of the associated nodes.

[0144] Assume that the system is evaluating the metadata risk of a viral genome scientific dataset. During public health emergencies such as an epidemic outbreak, the update frequency of the dataset and related management behaviors may increase significantly. For example, the data update frequency increases significantly, and more management feedback and quality control information are collected. According to the rule trigger mechanism, there is a correlation between the data update frequency of node 1 and the data integrity of node 2. Since the data update frequency increases significantly, the assessment task related to data integrity will be triggered.

[0145] In this embodiment, the latest update frequency data can be obtained by monitoring the metadata management behavior of similar data sets. For example, assume that the score of node 1 increases from 0.5 to 0.8. The system updates the score of node 2 using a weighting factor (e.g., α=0.6, β=0.4), combines the initial score of node 2 with the new score of node 1, and updates the risk score of node 2. This process is repeated until the node score reaches a preset convergence standard. The preset convergence standard is a set of indicators used to determine whether the risk score update has reached a stable state, specifically including a score change threshold, a relative change threshold, and an iteration limit. When the risk score of a node changes by less than a set threshold in several consecutive iterations, or does not reach the convergence standard within the specified number of iterations, the risk score of the node is judged to be converged to ensure the reliability and real-time nature of the risk assessment results. In this way, the system can dynamically adjust the risk assessment to ensure that during emergencies, the assessment results always reflect the latest data management behavior and risk status.

[0146] S53. Summarize the risk characteristics and risk scores of each node to obtain a risk score matrix for the assessment object.

[0147] After obtaining the final risk score of each node, the risk characteristics and risk scores of each node are summarized to form a more comprehensive risk score matrix for the assessment object.

[0148] In some optional embodiments, the risk score matrix may include feature names and risk scores, and may also include information such as the importance weight of each metadata feature.

[0149] like Figure 8 As shown, the metadata risk assessment method in this embodiment may also include: S61. Determine the risk level and risk influencing factors of the assessment object based on the risk assessment rules and the risk score matrix of the assessment object; S62. Generate a risk assessment report using a large language model based on risk levels and risk influencing factors.

[0150] The technical solution of the metadata risk assessment method for scientific datasets provided in this application is described below in conjunction with a specific embodiment.

[0151] like Fig. 9 As shown, the embodiment of the present application further provides a metadata risk assessment device 610 for a scientific data set. Optionally, the metadata risk assessment device 610 for a scientific data set may include: A task tree construction module 611 is used to construct a risk assessment task tree based on metadata features and preset risk assessment rules; A first risk indicator matrix acquisition module 612, used to acquire a first risk indicator matrix based on historical metadata of similar metadata; A second risk indicator matrix acquisition module 613, used to acquire a second risk indicator matrix based on the metadata management behavior of similar metadata; The risk assessment result generating module 614 is used to perform risk assessment on the scientific data set based on the risk assessment task tree and the risk indicator matrix to generate a risk assessment result.

[0152] Optionally, the device further comprises: A real-time acquisition module 615 is used to automatically acquire relevant information of metadata management behavior, metadata management behavior timing logic and behavior feature logic specificity; The result feedback module 616 is used to obtain feedback from users associated with the scientific data set on the risk assessment results, obtain the associated metadata features and risk assessment rule features based on the feedback information features, and optimize the risk factor weights and risk reference values.

[0153] In this embodiment, the risk assessment results can be classified according to metadata type and metadata feature clustering, the semantic features of the risk assessment results and the corresponding assessment feedback can be extracted, and a multi-classification risk feature knowledge base can be constructed to optimize the identification of risk factors and assessment dependencies.

[0154] Behavior timing logic refers to the logic that analyzes and describes the temporal order of a series of metadata management behaviors, which may include the time order. For example, whether a data sharing behavior occurs after data encryption, whether a data deletion operation is performed after data backup is completed, etc. Whether the time interval between the data update operation and the data verification operation exceeds the set threshold. By analyzing the behavior timing logic, the potential time-dependent risks of metadata can be identified. For example, if a data sharing operation is performed before data encryption is completed, it may bring the risk of data leakage, and the metadata has a time-dependent risk.

[0155] Behavior feature logical specificity refers to analyzing the feature differences or specificity of metadata management behaviors, which make some behaviors different from other behaviors in specific situations. It focuses on the feature differences of behaviors in different situations or by different executors, such as behavior mode, execution conditions, and metadata types involved.

[0156] In the specific implementation process, features can be extracted from the feedback information submitted by users. These features can be specific metadata fields, user opinions or suggestions on the evaluation results. For example, in the user feedback, "the frequency of data collection is too low, which affects the analysis results", the "data collection frequency" can be extracted as a metadata feature, and the risk assessment rules related to "data quality" can be identified, and then the extracted feedback information can be matched with the known metadata features and risk assessment rules. For example, if the feedback involves "data collection frequency", the system will automatically associate the extracted feedback information with the evaluation rules related to the feature, such as "data update timeliness" or "data integrity". This process relies on the system's built-in rules and feature mapping tables. In the specific implementation process, it can be implemented through machine learning models or predefined rule libraries.

[0157] In the process of optimizing the risk assessment mechanism, the weight distribution of risk factors can be flexibly adjusted according to the feedback information and the associated metadata features to enhance the accuracy of the assessment. Specifically, if the feedback frequently points out that the "frequency of data collection" has a significant impact on the accuracy of the analysis, the risk factor weight of the metadata feature is increased so that the risk factor has a greater influence in future assessments. In this embodiment, this can be achieved by retraining the risk assessment model or dynamically adjusting the rule weights.

[0158] In addition, risk reference values ​​can be adjusted based on feedback information. For example, if multiple feedback indicates that a particular encryption method is no longer secure, the corresponding risk reference value is updated to increase the requirements for new encryption standards. Updating the reference value ensures that the system uses the latest and most appropriate benchmarks to determine the risk level of data in future risk assessments.

[0159] The knowledge base construction module 617 is used to classify the risk assessment results according to metadata types and metadata feature clustering, extract semantic features of the risk assessment results and corresponding assessment feedback, and construct a multi-classification risk feature knowledge base.

[0160] The risk assessment task tree acquisition module 618 is used to acquire a risk assessment task tree based on metadata features and preset risk assessment rules.

[0161] It can be understood that the above-mentioned device embodiments and the above-mentioned method embodiments can correspond to each other, and similar descriptions of the device embodiments can refer to the method embodiments. To avoid repetition, it will not be repeated here. A metadata risk assessment device for a scientific data set provided in an embodiment of the present application can execute a metadata risk assessment method for a scientific data set provided in any embodiment of the present application, and has functional modules and beneficial effects corresponding to the execution method. The functional modules of the metadata risk assessment device for the scientific data set can be implemented in the form of hardware, can be implemented by instructions in the form of software, or can be implemented by a combination of hardware and software modules.

[0162] Specifically, each step of the method embodiment of the present application can be completed by the hardware integrated logic circuit and / or software instructions in the processor. The steps of the metadata risk assessment method of the scientific data set in the embodiment of the present application can be directly embodied as a hardware coding processor, or a combination of hardware and software modules in the coding processor. Optionally, the software module can be located in a random access memory, a read-only memory, a programmable read-only memory, a flash memory, an electrically erasable programmable memory, a register, and other storage media. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps in the above method embodiment in combination with its hardware.

[0163] The present application embodiment provides an electronic device 710, whose structure is as follows: Fig.10 The electronic device 710 may be Figure 1 The server 100 or the terminal 200 is shown.

[0164] like Fig.10 As shown, the electronic device 710 includes a memory 711, a processor 712, a communication module 713 and an input / output interface 714, etc. Optionally, the memory 711, the processor 712, the communication module 713 and the input / output interface 714 can be connected and communicated through a bus 715.

[0165] The memory 711 is used to store one or more computer programs and transmit the code of the computer program to the processor 712; when the one or more computer programs are executed by the processor 711, the metadata risk assessment method for scientific data sets in the embodiment of the present application is implemented.

[0166] Optionally, the electronic device 710 can be connected to a network via a communication module 713 to communicate with other devices, such as a terminal or a server, through the network to achieve data interaction. The electronic device 710 can be various forms of digital computers, such as desktop computers, servers, workstations, mainframe computers or other types of computers. The electronic device 710 can also be various forms of mobile terminals, such as smart phones, tablet computers, wearable devices (such as helmets, glasses, watches, etc.) and other similar mobile terminals.

[0167] Optionally, the electronic device 710 can be connected to the required input / output devices, such as a keyboard, a display device, etc., through the input / output interface 714. The electronic device 710 itself can have a display device, and can also be connected to other display devices through the input / output interface 714. Optionally, a storage device, such as a hard disk, can also be connected through the input / output interface 714, so that data in the electronic device 710 can be stored in the storage device, or data in the storage device can be read, and data in the storage device can also be stored in the memory 711. It can be understood that the input / output interface 714 can be a wired interface or a wireless interface. Depending on the actual application scenario, the device connected to the input / output interface 714 can be a component of the electronic device 710, or it can be an external device connected to the electronic device 710 when needed.

[0168] Optionally, the memory 711 may be a volatile memory and / or a non-volatile memory, the volatile memory may be a random access memory, etc., and the non-volatile memory may be a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory or a flash memory, etc.

[0169] Optionally, the computer program stored in the processor 711 may be divided into one or more modules, which are stored in the memory 711 and executed by the processor 712 to complete the method provided in the embodiment. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the computer program instruction segments are used to describe the execution process of the computer program in the electronic device 710.

[0170] Optionally, the processor 712 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 712 include, but are not limited to, a central processing unit, a graphics processing unit, a digital signal processor, various special artificial intelligence computing chips, various processors that run machine learning model algorithms, and may also be any appropriate controller, microcontroller, processor, etc. The processor 712 executes various methods and processes of this embodiment, illustratively, such as a metadata risk assessment method for a scientific data set in an embodiment of the present application.

[0171] Optionally, the bus 715 may include a path for transmitting information. The bus 715 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. According to different functions, the bus 715 may be divided into an address bus, a data bus, a control bus, etc.

[0172] Obviously, the above embodiments of the present application are merely examples for clearly illustrating the technical solution of the present application, and are not intended to limit the specific implementation methods of the present application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the claims of the present application shall be included in the protection scope of the claims of the present application.

Claims

1. A metadata risk assessment method for scientific datasets, characterized in that: The method comprises: Get metadata features and similar metadata of scientific datasets; Obtain a risk assessment task tree based on metadata features and preset risk assessment rules; Based on the historical metadata of similar metadata, a first risk indicator matrix is ​​obtained; Based on the metadata management behavior of similar metadata, a second risk indicator matrix is ​​obtained; Based on the risk assessment task tree and the first risk indicator matrix and the second risk indicator matrix, a risk assessment is performed on the scientific data set to generate a risk assessment result.

2. A metadata risk assessment method for scientific data sets according to claim 1, characterized in that: The step of obtaining a risk assessment task tree based on metadata features and preset risk assessment rules specifically includes: Identify risk factors and assessment dependencies in metadata features based on risk assessment rules; Determine the risk assessment level based on the mapping relationship between risk assessment rules and metadata features; Constructing an evaluation data-aware network based on evaluation dependencies; A risk assessment task tree is established based on metadata features, risk factors, assessment dependencies, risk assessment levels and data perception networks; the risk assessment task tree includes multiple nodes, and each node information includes metadata features, risk factors, and assessment dependencies.

3. A metadata risk assessment method for scientific data sets according to claim 2, characterized in that: The obtaining of the first risk indicator matrix based on the historical metadata of the similar metadata specifically includes: Acquire historical metadata of similar metadata, perform correlation analysis based on metadata feature changes of the historical metadata to obtain correlation coefficients between metadata features of the historical metadata, and construct a metadata feature association matrix based on the correlation coefficients; the metadata feature association matrix includes metadata associations and metadata type associations; According to metadata feature types, the latest metadata of similar metadata are respectively obtained, and cluster analysis is performed on the latest metadata to obtain reference values ​​of metadata of different types; the metadata feature types include one or more of descriptive features, usage features, and quality features; The similarities between metadata features of different types and metadata reference values ​​are calculated, and a first risk indicator matrix is ​​generated according to the correlation coefficients and the similarities.

4. The method for risk assessment of metadata of a scientific dataset according to claim 1, characterized in that: The metadata management behavior based on similar metadata, obtaining a second risk indicator matrix, specifically includes: Obtain the metadata management behavior semantics of similar metadata, and obtain the metadata management behavior features through semantic feature clustering analysis; The metadata features associated with the metadata management behavior features are obtained, and according to the mapping relationship between the metadata features and the risk assessment rules, a second risk indicator matrix containing the metadata management behavior risk scores is obtained.

5. The method for risk assessment of metadata of a scientific data set according to claim 2, characterized in that: The risk assessment of the scientific data set is performed based on the risk assessment task tree and the first risk indicator matrix and the second risk indicator matrix to obtain the risk score matrix of the assessment object, specifically including: Determine the risk reference value of each node according to the metadata characteristics of each node of the risk assessment task tree and the first risk indicator matrix and the second risk indicator matrix; Calculate the risk score of each node based on the risk reference value and risk factor of each node; The metadata characteristics and risk scores of each node are summarized to obtain the risk score matrix of the assessment object.

6. A metadata risk assessment method for scientific data sets according to claim 5, characterized in that: The risk score of each node is calculated based on the risk reference value and risk factor of each node, specifically including: Based on the mapping relationship between metadata features and risk assessment rules, the corresponding risk factor weights are determined; Extracting a first subset matrix related to the metadata features from the first risk indicator matrix, and weighting the associated risk factors based on the correlation coefficients between the metadata features of the first subset matrix to obtain a first weighted risk factor; Extracting a second subset matrix related to the metadata management behavior from the second risk indicator matrix, and weighting the associated risk factors based on the metadata management behavior risk score of the second subset matrix to obtain a second weighted risk factor; Calculating a risk score for each node based on the risk reference value of each node, the first weighted risk factor, and the second weighted risk factor; Based on the evaluation dependency of each node, the associated nodes are obtained, and the risk scores of the associated nodes are calculated. The risk scores of the associated nodes are iteratively updated based on the evaluation data perception network to obtain the final risk scores of each node.

7. The method for risk assessment of metadata of a scientific dataset according to claim 1, characterized in that: The metadata risk assessment method further includes: Determine the risk level and risk influencing factors of the assessment object based on risk assessment rules and the risk score matrix of the assessment object; Based on risk levels and risk influencing factors, a risk assessment report is generated using a large language model.

8. A metadata risk assessment device for a scientific data set, characterized in that: The device comprises: The metadata acquisition module is used to obtain metadata features and similar metadata of scientific datasets; A task tree construction module is used to construct a risk assessment task tree based on metadata features and preset risk assessment rules; A first risk indicator matrix acquisition module, used to acquire a first risk indicator matrix based on historical metadata of similar metadata; A second risk indicator matrix acquisition module, used for acquiring a second risk indicator matrix based on metadata management behavior of similar metadata; The risk assessment result generation module is used to perform risk assessment on scientific data sets based on the risk assessment task tree and risk indicator matrix and generate risk assessment results.

9. The metadata risk assessment device for scientific data sets according to claim 8, characterized in that: The device also includes: Real-time acquisition module, used to automatically obtain relevant information of metadata management behavior, metadata management behavior timing logic and behavior feature logic specificity; The result feedback module is used to obtain feedback from users associated with scientific data sets on risk assessment results, obtain associated metadata features and risk assessment rule features based on feedback information features, and optimize risk factor weights and risk reference values; The knowledge base construction module is used to classify risk assessment results according to metadata type and metadata feature clustering, extract semantic features of risk assessment results and corresponding assessment feedback, and build a multi-classification risk feature knowledge base.

10. An electronic device, characterized in that: include: a memory for storing one or more computer programs; A processor, when the one or more computer programs are executed by the processor, implements the metadata risk assessment method for a scientific dataset as described in any one of claims 1 to 7.

Images