A data security compliance evaluation and identification method and system based on knowledge reasoning

Abstract

A data security compliance evaluation and identification method and system based on knowledge reasoning, characterized in that the method comprises the following steps: collecting legal regulation texts and constructing a regulation knowledge graph; performing entity recognition on reported log texts or compliance point query texts through the regulation knowledge graph to obtain entity ontology and entity events; setting sub-compliance indicators, fusing the entity ontology, entity events and log texts, inputting the fused features into the regulation knowledge graph to classify the fused features into the sub-compliance indicators, thereby calculating the compliance scores of the log texts for each indicator and judging the compliance levels of behaviors; and constructing a question and answer system to obtain compliance points in the regulation knowledge graph through compliance point query texts. The present application is objective, time-saving, accurate, fully understands compliance requirements, and has high flexibility and adaptability.

Description

Technical Field

[0001] This invention relates to the field of data compliance, and more specifically, to a data security compliance assessment and identification method and system based on knowledge reasoning. Background Technology

[0002] In today's digital age, power grid companies and their subsidiaries widely utilize online partnership channels for marketing activities. However, with increasing compliance requirements and regulatory focus on data security and privacy protection, these multi-business groups need to ensure they adhere to all applicable compliance requirements in their online partnerships. Identifying compliance requirements is a critical task, involving the understanding and interpretation of a large number of rules, policies, contracts, and industry standards.

[0003] Currently, many companies still rely on manual identification and verification of compliance requirements. This method suffers from subjectivity and time-consuming issues, is prone to omissions or errors in identifying compliance requirements, and cannot meet the needs of large-scale and efficient compliance. Some automated methods use keyword matching to identify compliance requirements. However, this method suffers from semantic ambiguity and insufficient contextual understanding. It can only match based on predefined keywords and cannot handle the complexity and diversity of compliance requirements.

[0004] Within power grid companies, as business operations expand, different systems, platforms, and applications are implemented using various technical means. The complexity and diversity of these implementation processes make it difficult to simply assess the security and compliance attributes of these technical means.

[0005] Human methods for evaluating large-scale integrated systems and complex applications are subject to numerous complex factors. Existing methods suffer from drawbacks such as subjectivity, time consumption, semantic ambiguity, insufficient contextual understanding, lack of flexibility and adaptability.

[0006] On the other hand, to date, there is still no effective automated means to accurately and effectively assess the security compliance of various power grid operations. For example, log information contained in operational logs can be used to measure and assess the security compliance of various systems or applications, but only experienced technical personnel can accurately interpret the contents of these logs. Furthermore, if the security compliance indicators stipulated in various laws and regulations are not understood, even if the log content can be parsed, the compliance level of the application still cannot be judged. In addition, due to the large amount of content, volume, and difficulty in interpretation and accurate understanding of legal and regulatory documents, automated judgment methods do not yet exist.

[0007] To address the aforementioned issues, there is an urgent need for a data security compliance assessment and identification method and system. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a data security compliance key point identification method based on knowledge reasoning. By collecting legal and regulatory texts, constructing a legal knowledge graph, performing entity matching and identification on the reported log texts and knowledge graph to obtain entity subjects and entity events, setting compliance indicators, classifying texts to corresponding indicators and summarizing them into compliance scores, and constructing an intelligent question-answering system to help deduce compliance key points in laws and regulations.

[0009] The present invention adopts the following technical solution.

[0010] The first aspect of this invention relates to a data security compliance assessment and identification method based on knowledge reasoning. The method includes the following steps: collecting legal and regulatory texts and constructing a regulatory knowledge graph; identifying entities from reported log texts or compliance point query texts using the regulatory knowledge graph to obtain entity ontology and entity events; setting subdivided compliance indicators, fusing features of entity ontology, entity events, and log texts, inputting the fused features into the regulatory knowledge graph to classify the fused features into subdivided compliance indicators, thereby calculating the compliance score of the log text for each indicator and judging the compliance level of the behavior; constructing a question-and-answer system; and obtaining compliance points from the regulatory knowledge graph by querying texts in the question-and-answer system using compliance points.

[0011] Preferably, constructing a legal knowledge graph also includes: classifying entity nodes into three categories based on node attributes: legal clause nodes, legal concept nodes, and legal entity nodes; classifying node relationships into three categories based on edge attributes: inclusion relationships, involvement relationships, and penalty relationships; inclusion relationships are used to establish the association between legal concept nodes and legal entity nodes, involvement relationships are used to establish the association between legal clause nodes and legal concept nodes, and penalty relationships are used to establish the association between legal clauses and penalty-type legal entities.

[0012] Preferably, constructing a legal knowledge graph further includes: determining the method for extracting entity nodes based on the length of the legal text; for legal texts shorter than a preset threshold, manually annotating the entity nodes in the legal text and defining the attributes of the entity nodes as legal clauses, legal concepts, or legal entities, thereby extracting nodes and relationships in the legal text and constructing triples to achieve the initial construction of the legal knowledge graph; for legal texts longer than a preset threshold, using natural language processing methods to parse the legal text, obtaining the nodes and relationships in the legal text, constructing triples, and updating the initially constructed legal knowledge graph.

[0013] Preferably, the detailed compliance indicators include primary scenario indicators and secondary specific indicators; among them, the primary scenario indicators are classified according to the application scenario of the log text to be evaluated and the normative objectives of laws and regulations; the secondary specific indicators include the scope of data use, the timeliness of data use, and the purpose and method of information application.

[0014] Preferably, the primary scenario indicators include the privacy policy for marketing business application scenarios; the secondary scenario indicators under the privacy policy for marketing business application scenarios include the scope of data use, the timeliness of data use, and the purpose and method of information application.

[0015] Preferably, the compliance score for the primary scenario indicator is:

[0016]

[0017] Among them, C i This represents the score of the primary scenario indicator on the i-th secondary evaluation indicator, with a value of 0 or 1.

[0018] n represents the number of secondary evaluation indicators.

[0019] Preferably, text segmentation is obtained, and the text segmentation is combined with entity ontology and entity events to form the fused features; the fused features are input into a pre-trained BERT model to classify the fused features and obtain the classification results of the fused features.

[0020] Preferably, based on the classification results of the fusion features, the log text features are classified into different indicators, the scores of different indicators are obtained and summarized to obtain a compliance score; the compliance score is then mapped to the corresponding compliance level to determine the level of compliance.

[0021] A second aspect of this invention relates to a data security compliance assessment and identification system based on knowledge reasoning, utilizing the method of the first aspect of this invention. The system includes a graph construction module, an identification module, a calculation module, a system construction module, and a query module. The graph construction module collects legal and regulatory texts to construct a legal knowledge graph. The identification module identifies entities from reported log texts or compliance point query texts using the legal knowledge graph, obtaining entity ontologies and entity events. The calculation module sets detailed compliance indicators, fuses the entity ontologies, entity events, and log texts for feature analysis, inputs the fused features into the legal knowledge graph, classifies the fused features into detailed compliance indicators, calculates the compliance score for each indicator, and assesses the compliance level of the behavior. The system construction module constructs a question-and-answer system. The query module retrieves compliance points from the legal knowledge graph by querying texts from the question-and-answer system.

[0022] A third aspect of the present invention relates to a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to perform operations according to the instructions to execute the steps of the method in the first aspect of the present invention.

[0023] A fourth aspect of the present invention relates to a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method of the first aspect of the present invention.

[0024] The beneficial effects of this invention are that, compared with the prior art, the data security compliance key point identification method based on knowledge reasoning in this invention can collect legal and regulatory texts, construct a legal knowledge graph, perform entity matching and identification on the reported log texts and knowledge graph to obtain entity subjects and entity events, set compliance indicators, classify texts to corresponding indicators and summarize them into compliance scores, and build an intelligent question-answering system to help deduce the compliance key points in laws and regulations. This invention is objective and time-saving, provides accurate results, fully understands the compliance requirements in laws and regulations, and has high flexibility and adaptability for different data businesses.

[0025] The beneficial effects of the present invention also include:

[0026] 1. The method uses knowledge reasoning, combined with technologies such as knowledge graphs, natural language processing, and machine learning, to solve the problem of identifying and detecting compliance points in business data. It can quickly and accurately obtain compliance scores in various online external cooperation channels for power marketing and identify key compliance points.

[0027] 2. The method predefines multi-level evaluation indicators and uses machine learning models to accurately classify the input text, thereby achieving compliance ratings for different indicators and ensuring the accuracy of the rating process.

[0028] 3. Method: Before inputting the input text into the BERT model for training, the features of the input text were fully expanded, and enhanced features were used for training, which fully ensured the richness of the training data. Attached Figure Description

[0029] Figure 1 This is a schematic diagram illustrating the steps of a data security compliance assessment and identification method based on knowledge reasoning according to the present invention;

[0030] Figure 2 This is a schematic diagram illustrating the construction method of entities and relationships in a regulatory knowledge graph in a data security compliance assessment and identification method based on knowledge reasoning according to the present invention.

[0031] Figure 3 This is a schematic diagram of the compliance level assessment and compliance requirement query process in a data security compliance assessment and identification method based on knowledge reasoning according to the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this invention are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments not described in this invention obtained by those skilled in the art based on the embodiments described in this invention without creative effort should fall within the protection scope of this invention.

[0033] Figure 1 This is a schematic diagram illustrating the steps of a data security compliance assessment and identification method based on knowledge reasoning according to the present invention. The present invention relates to a method for identifying key points of data security compliance based on knowledge reasoning, such as... Figure 1 As shown, it includes the following steps:

[0034] Step 1: Collect legal and regulatory texts and construct a legal knowledge graph.

[0035] In step 1:

[0036] The construction of a regulatory knowledge graph involves organizing professional terminology texts into knowledge. By using a regulatory knowledge graph, straightforward and easy-to-understand texts can be transformed into professional and targeted regulatory texts, which can enhance compliance detection and the identification of compliance points.

[0037] A knowledge graph is a structured graphical representation used to store and organize key concepts, rules, and relationships within legal texts. To construct a legal knowledge graph, the method collects legal texts related to marketing operations, such as the Cybersecurity Law, including relevant laws and regulations, policy documents, contract templates, and industry standards. These legal texts undergo preprocessing, including text cleaning, word segmentation, and entity recognition, to extract key information and concepts.

[0038] Preferably, constructing a regulatory knowledge graph also includes:

[0039] Based on their node attributes, entity nodes are divided into three categories: legal clause nodes, legal concept nodes, and legal entity nodes.

[0040] Node relationships are categorized into three types based on edge attributes: containment relationships, involvement relationships, and penalty relationships.

[0041] Among them, the inclusion relationship is used to establish the association between legal concept nodes and legal entity nodes, the involvement relationship is used to establish the association between legal clause nodes and legal concept nodes, and the penalty relationship is used to establish the association between legal clauses and penalty-type legal entities.

[0042] In this invention, nodes and edges of a legal knowledge graph are established based on the semantic relationships and logical structure of legal texts. Nodes are divided into three categories: legal clauses, legal concepts, and legal entities.

[0043] Legal provisions are explicit and specific clauses in laws and regulations. Legal concepts refer to general principles, concepts, or theories within the legal system that help explain and guide the application of legal provisions. In cybersecurity law, legal concepts can relate to types of illegal events and types of illegal subjects. Legal entities represent the subordinate concepts or specific entities corresponding to legal concepts.

[0044] Edges represent relationships between different entities. For example, the inclusion relationship describes the relationship between a legal concept and a legal entity, i.e., an "is-a" relationship. The involvement relationship describes the relationship between a legal clause and a legal concept, i.e., which legal concepts are involved in a particular legal clause. The penalty relationship describes the relationship between a prescriptive legal clause and a punitive legal entity, i.e., which punitive clause applies when a prescriptive clause is violated.

[0045] Figure 2 This is a schematic diagram illustrating the construction method of entities and relationships in the regulatory knowledge graph in the data security compliance assessment and identification method based on knowledge reasoning of the present invention. If... Figure 2 As shown, preferably, constructing a regulatory knowledge graph also includes:

[0046] Based on the length of the legal and regulatory texts, the method for extracting entity nodes is determined;

[0047] For legal and regulatory texts shorter than a preset threshold, entity nodes in the legal and regulatory texts are manually annotated, and the attributes of the entity nodes are defined as legal clauses, legal concepts, or legal entities, thereby extracting nodes and relationships in the legal and regulatory texts and constructing triples to achieve the initial construction of a legal knowledge graph.

[0048] For legal and regulatory texts that exceed a preset threshold in length, natural language processing methods are used to parse the text, obtain the nodes and relationships within it, construct triples, and update the initially constructed legal and regulatory knowledge graph.

[0049] A semi-automated approach is used to construct the legal knowledge graph. For shorter regulations such as the Cybersecurity Law, legal concepts and clauses are extracted manually. For longer regulations, natural language processing semantic analysis techniques are used to perform semantic parsing and information extraction. By identifying and extracting keywords, terms, definitions, obligations, and prohibited behaviors from the legal text, these are mapped to the corresponding nodes and edges in the legal knowledge graph.

[0050] In constructing the legal knowledge graph, we also considered the timeliness and variability of legal texts. The knowledge graph is regularly updated and maintained to ensure that the legal information and relationships within it remain up-to-date and accurate.

[0051] Figure 3 This is a schematic diagram illustrating the process of compliance level assessment and compliance requirement query in a data security compliance assessment and identification method based on knowledge reasoning according to the present invention. Figure 3 As shown, this invention includes two main functions: compliance level assessment and compliance requirement query. In the specific implementation process, log text and compliance point query text are used respectively, and entity recognition is required for both types of text first.

[0052] Step 2: Use the legal knowledge graph to identify entities and events from the reported log text or compliance key point query text.

[0053] In step 2:

[0054] Log text is textual information generated by an organization or system that records various operations, events, or interactions. This text can contain user actions, network activity, server events, etc. The format is typically structured or semi-structured, such as JSON, XML, CSV, or text files, and may contain data items such as timestamps, event descriptions, IP addresses, user identifiers, and operation types. It is usually automatically generated by the system, application, device, or service.

[0055] Query text refers to text queries provided by users or the system to find compliance points or specific information. This can be a natural language question, a keyword, a question statement, etc. Query text is typically in natural language format and is provided by the system administrator or other authorized personnel. In this invention, log text serves as the primary basis for determining whether a system or application is compliant and its compliance level. Query text, on the other hand, can be used to search for specific compliance requirements in laws and regulations, facilitating developers in improving the security and compliance of the system or application.

[0056] The original log text data and compliance requirement query statements are used to obtain the corresponding entity events and entity ontology in the constructed regulatory knowledge graph through the following steps, which will be used for subsequent compliance level classification and compliance point query.

[0057] First, the input log text or query statement needs to be cleaned and preprocessed to filter out sentences with relevant entities. This includes removing special characters, punctuation marks, and unnecessary whitespace from the text, and segmenting the text into sentences and words.

[0058] Events are typically composed of phrases consisting of verbs and nouns, such as "data needs to be anonymized." First, part-of-speech tagging (POT) techniques are used to extract key phrases that match the POT's part of speech. Then, based on the tagged POT sequence, rule-based matching methods are used to identify and extract the event. Similarly, entity extraction can be performed by matching entities based on their initial POT tag and other specific POT patterns. For example, consecutive noun POT tags can represent a noun phrase, which could be an entity such as a person's name or a place name.

[0059] After entity extraction, a set of key entities is obtained. Related entities for each key entity in the knowledge graph are then retrieved, and a candidate entity set D = {d1, d2, ..., dn} is constructed. n}, d i It retrieves the descriptive text of each entity from the candidate entity set, which can include the entity's definition, attributes, relationships, and other related information.

[0060] To obtain contextual information about key entities, a sliding window of length l can be set up to extract l words before and after the key entity as contextual information.

[0061] The contextual information of the key entity is compared with the textual description information of each entity in the candidate entity description text set D. The calculation method uses the cosine similarity formula:

[0062]

[0063] Where V c V is a vector representing the contextual information of key entities. d Sim(V) represents a vector of candidate entity description text. c V d The similarity between two vectors is denoted by .

[0064] Based on the text similarity calculation results, the entity with the highest similarity is selected as the final entity ontology mapping result. A similarity threshold λ is set, and only entities with a similarity exceeding the threshold λ are mapped.

[0065] Finally, the log text is mapped to the pre-built regulatory knowledge graph to obtain entity ontology and event entity, which are used for subsequent compliance determination and identification of compliance points.

[0066] Step 3: Set up detailed compliance indicators, fuse the entity ontology, entity events and log text for feature analysis, and input the fused features into the regulatory knowledge graph to classify the fused features into the detailed compliance indicators, thereby calculating the compliance score of the log text for each indicator and judging the compliance level of the behavior.

[0067] Furthermore, a question-and-answer system is constructed to retrieve compliance points from the regulatory knowledge graph through the compliance point query text.

[0068] In step 3:

[0069] Compliance metrics are key measures used to assess and monitor an organization’s compliance with specific laws or policies. The content and structure of these metrics can vary depending on the specific area of ​​compliance requirements and the organization’s needs.

[0070] Preferably, the detailed compliance indicators include primary scenario indicators and secondary specific indicators; wherein,

[0071] The primary scenario indicators are classified based on the application scenario of the log text to be evaluated and the normative objectives of laws and regulations.

[0072] In one embodiment of the present invention, a privacy policy for a marketing business application scenario is used as a primary scenario indicator. Other indicators may also include different business application scenarios. A privacy policy is a regulatory objective for the security compliance of business deployments under a specific law or regulation. For example, in the Cybersecurity Law, regulatory objectives may include network operation security, network information security, monitoring and early warning, and emergency response.

[0073] The specific secondary indicators are based on the scope of data use, the timeliness of data use, and the purpose and method of information application.

[0074] In summary, in this invention, the primary scenario is the main category or area of ​​compliance requirements, while the secondary scenario is a subcategory or specific content of the primary scenario.

[0075] As is easily conceivable, the scoring rules for each indicator are formulated simultaneously with the subdivided compliance indicators. The classification model described below obtains the characteristics of the log text under each indicator, which serves as the scoring basis. Finally, the method yields the total score for each indicator. For example, when the primary indicator is privacy policy, and the secondary indicator is the scope of data use, the compliance statement refers to whether it falls within the data use scope of the relevant department or position. Specifically, this is achieved by binary classifying the log text according to its compliance statement using the following model, thereby obtaining the corresponding compliance points. Finally, the scores of all indicators are summed to quantify the degree of compliance.

[0076] In summary, by fusing features from the obtained knowledge graph corresponding to entity events, entity ontologies, and log text, and then classifying them into the corresponding compliance requirement indicators through a classification model, the compliance level is finally obtained.

[0077] Specifically, different indicators are designed based on different data scenarios. First-level scenario indicators and second-level specific indicators are set. A compliance score is given based on whether the second-level specific indicators are completed; completion is recorded as 1, and non-completion as 0. The compliance score is obtained using the following formula.

[0078]

[0079] Among them, C i The score represents the privacy policy's score on the i-th evaluation indicator, where n is the number of secondary evaluation indicators. The indicator design includes, for example, a primary indicator for privacy policies in marketing business application scenarios, and corresponding secondary indicators for data usage scope, data usage timeliness, and the purpose and method of information application. The score indicates whether the corresponding indicator meets the stipulated requirements.

[0080] Preferably, the text segmentation {a1, a2, ..., a...} is obtained. n-1 ,a n}, combining text segmentation with entity ontology and physical events The fused features are then input into a pre-trained BERT model to classify the fused features.

[0081] Entities are obtained through knowledge graphs, then fused with text segmentation for enhancement before being fed into the model for classification. This method achieves better classification results than directly classifying entities without knowledge graph extraction. The method first obtains text segmentation {a1, a2, ..., a...} n-1 ,a n}. Where a1,…,a n All are individual words obtained after text segmentation. Simultaneously, the entity ontology for entity recognition is obtained based on step 2. and physical events New text feature representations can be obtained.

[0082] Then, the pre-trained BERT model is used to obtain the word vector representations of all words in the legal knowledge graph for this case. Some entities are represented using phrases or longer word groups, which need to be segmented first, and then the average of the segmented word vectors is used as the vector representation of the entity. The calculated sentence vectors are input into the CNN layer for training. In one embodiment of the present invention, three convolutional kernels with channel sizes of 2×768, 3×768, and 4×768 are used. To avoid gradient explosion and improve the training speed of the network, we perform ReLU activation function operation on the text vectors using the convolutional kernels. After the ReLU operation, we perform feature fusion of the text vectors and connect the feature-fused text vectors to the Softmax activation function through a fully connected layer for classification. The Softmax calculation formula is as follows.

[0083]

[0084] P i Let e ​​represent the probability of class i. i Let j represent the corresponding value of the output for the i-th category, and j represent the total number of categories. The loss function formula is as follows:

[0085]

[0086] in Let y be the prediction vector, y be the true classification value, and α be the penalty weight vector.

[0087] Finally, the log text features are categorized into different metrics, scores for each metric are obtained and aggregated, resulting in a final compliance score. This compliance score can then be mapped to a corresponding compliance level, allowing for the determination of the level of compliance.

[0088] Build a question-and-answer system to help deduce compliance points in the regulatory knowledge graph by querying compliance point texts.

[0089] By obtaining the entity of the statement in the regulatory knowledge graph through the user's query statement, and performing reasoning in the knowledge graph, the corresponding compliance points can be obtained.

[0090] First, a method is used to transform the user's query for compliance requirements at a specific stage into a form that the system can understand. This is achieved by obtaining the corresponding entities through entity recognition.

[0091] Secondly, the definition of inference rules is based on known entities and relationships, defining a series of inference rules to describe how to reason using existing information. Inference rules can be based on logical rules, statistical rules, rule expressions, etc. The following rules are defined:

[0092] If a triple A-relation-B exists, and A is known, then B can be deduced.

[0093] If there exists a triple A-relation-B, and B is known, then A can be deduced.

[0094] Using known entities and relationships as query conditions, the system searches the knowledge graph for matching triples. This provides further information for inference. Based on the known entities, relationships, and query results, logical reasoning is performed using inference rules to deduce the target entity. The inference chain is progressively expanded by applying inference rules, passing information from known entities and relationships to new entities. During the inference process, multiple candidate entities may conform to the inference rules. Candidate entities are evaluated and selected based on factors such as credibility and relevance. Indicators such as probability, confidence, and similarity are used to measure the likelihood of candidate entities.

[0095] Finally, based on the reasoning results, an answer matching the user's query is generated. Using the obtained entity information, the final information is returned as the answer from the question-and-answer system, allowing the user to obtain the corresponding information and thus understand the legal and regulatory compliance requirements for the specified stage.

[0096] The second aspect of this invention relates to a data security compliance assessment and identification system based on knowledge reasoning, utilizing the method of the first aspect of this invention. The system includes a graph construction module, an identification module, a calculation module, a system construction module, and a query module. The graph construction module collects legal and regulatory texts to construct a legal knowledge graph. The identification module identifies entities from reported log texts or compliance point query texts using the legal knowledge graph, obtaining entity ontologies and entity events. The calculation module sets detailed compliance indicators, fuses the entity ontologies, entity events, and log texts for feature analysis, inputs the fused features into the legal knowledge graph, classifies the fused features into detailed compliance indicators, calculates the compliance score for each indicator, and assesses the compliance level of the behavior. The system construction module constructs a question-and-answer system, and the query module retrieves compliance points from the legal knowledge graph by querying texts from the question-and-answer system.

[0097] A third aspect of the present invention relates to a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to perform operations according to the instructions to execute the steps of the method in the first aspect of the present invention.

[0098] A fourth aspect of the present invention relates to a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method of the first aspect of the present invention.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A data security compliance assessment and identification method based on knowledge reasoning, characterized in that, The method includes the following steps: Collect legal and regulatory texts and construct a legal knowledge graph; The reported log text or compliance key point query text is used to identify entities through the aforementioned regulatory knowledge graph to obtain entity ontology and entity events; The system sets up detailed compliance metrics, fuses the entity ontology, entity events, and log text for feature analysis, inputs the fused features into a pre-trained BERT model to classify the fused features, and then classifies the fused features into the detailed compliance metrics. For each metric, the system calculates the compliance score of the log text and evaluates the compliance level of the behavior. The detailed compliance indicators include primary scenario indicators and secondary specific indicators. The primary scenario indicators are classified based on the application scenario of the log text to be evaluated and the regulatory objectives of laws and regulations. The secondary specific indicators include the scope of data use, the timeliness of data use, and the purpose and method of information application. The primary scenario indicators include the privacy policy for marketing business application scenarios; the secondary scenario indicators under the privacy policy for marketing business application scenarios include the scope of data use, the timeliness of data use, and the purpose and method of information application. The compliance score for the first-level scenario indicator is as follows: in, The score of the primary scenario indicator on the i-th secondary evaluation indicator takes the value of 0 or 1; n is the number of secondary evaluation indicators; Build a question-and-answer system; By querying the text from the question-and-answer system, compliance points in the regulatory knowledge graph can be obtained.

2. The data security compliance assessment and identification method based on knowledge reasoning according to claim 1, characterized in that: The construction of the regulatory knowledge graph also includes: Based on their node attributes, entity nodes are divided into three categories: legal clause nodes, legal concept nodes, and legal entity nodes. Node relationships are categorized into three types based on edge attributes: containment relationships, involvement relationships, and penalty relationships. The inclusion relationship is used to establish the association between legal concept nodes and legal entity nodes; the involvement relationship is used to establish the association between legal clause nodes and legal concept nodes; and the penalty relationship is used to establish the association between legal clauses and penalty-type legal entities.

3. The data security compliance assessment and identification method based on knowledge reasoning according to claim 2, characterized in that: The construction of the regulatory knowledge graph also includes: Based on the length of the aforementioned legal and regulatory text, the method for extracting entity nodes is determined; For legal and regulatory texts with a length less than a preset threshold, entity nodes in the legal and regulatory texts are manually labeled, and the attributes of the entity nodes are defined as legal clauses, legal concepts, or legal entities, thereby extracting entity nodes and node relationships in the legal and regulatory texts and constructing triples to achieve the initial construction of a legal knowledge graph; For legal and regulatory texts that exceed a preset threshold in length, natural language processing methods are used to parse the legal and regulatory texts, obtain entity nodes and node relationships in the legal and regulatory texts, construct triples, and update the initially constructed legal and regulatory knowledge graph.

4. The data security compliance assessment and identification method based on knowledge reasoning according to claim 1, characterized in that: Obtain text segmentation, and combine the text segmentation with the entity ontology and entity events to form the fusion feature.

5. The data security compliance assessment and identification method based on knowledge reasoning according to claim 4, characterized in that: Based on the classification results of the fusion features, the log text features are classified into different indicators, the scores of different indicators are obtained and summarized to obtain the compliance score. The compliance score is assigned to a corresponding compliance level to determine the level of compliance.

6. A data security compliance assessment and identification system based on knowledge reasoning, utilizing the method described in any one of claims 1-5, characterized in that: The system includes a map construction module, a recognition module, a calculation module, and a system construction and query module; wherein, The graph construction module is used to collect legal and regulatory texts and construct a legal knowledge graph. The identification module is used to identify entities by passing the reported log text or compliance key point query text through the regulatory knowledge graph to obtain entity ontology and entity event; The calculation module is used to set detailed compliance indicators, fuse the entity ontology, entity events and log text for feature extraction, input the fused features into a pre-trained BERT model to classify the fused features, classify the fused features into the detailed compliance indicators, thereby calculating the compliance score of the log text for each indicator and judging the compliance level of the behavior. The detailed compliance indicators include primary scenario indicators and secondary specific indicators. The primary scenario indicators are classified based on the application scenario of the log text to be evaluated and the regulatory objectives of laws and regulations. The secondary specific indicators include the scope of data use, the timeliness of data use, and the purpose and method of information application. The primary scenario indicators include the privacy policy for marketing business application scenarios; the secondary scenario indicators under the privacy policy for marketing business application scenarios include the scope of data use, the timeliness of data use, and the purpose and method of information application. The compliance score for the first-level scenario indicator is as follows: in, The score of the primary scenario indicator on the i-th secondary evaluation indicator takes the value of 0 or 1; n is the number of secondary evaluation indicators; The system construction module is used to build a question-and-answer system; The query module retrieves compliance points from the regulatory knowledge graph by querying text in the question-and-answer system.

7. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-5.

Images