A method and system for secure access management of enterprise sensitive data
Through multi-dimensional analysis and dynamic evaluation, sensitivity index is generated, and adaptive access control is carried out in combination with user attributes and data characteristics, dynamic adjustment and encryption processing of data security access management methods are realized, and the rigidity of data security access management in the existing technology is solved. The problem improves the system's flexibility and complex attack resistance.
Patent Information
- Application Number
- CN202411126433.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-16
AI Technical Summary
The existing data security access management methods lack dynamic and context-awareness, cannot reflect changes in data value in a timely manner, the access control strategy is rigid, and it is difficult to adapt to complex collaboration scenarios and user behavior patterns, the encryption strategy affects the efficiency of data usage, and in abnormal behavior Deficiencies in detection and threat response.
Through multi-dimensional analysis and dynamic evaluation, sensitivity index is generated, correlation analysis is performed based on user attributes and data characteristics to realize adaptive access control; full-scale recording and pattern analysis of user behavior, security domain division and collaboration strategy configuration are carried out, and differences are made Integrate multi-source information for threat analysis, dynamically adjust defense policies and update access control matrix.
Realize dynamic adjustment of access control policies that reflect changes in data value in real time, improve data access flexibility and security, promote cross-departmental collaboration, optimize system performance, and enhance resistance to complex attacks, so that the system can optimize itself Adapt to changing safety environments.
Smart Images

Figure CN118656870B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of secure access to sensitive data, and in particular to a method and system for managing secure access to sensitive data of an enterprise. Background Art
[0002] As digital transformation deepens, the value and importance of enterprise data are becoming increasingly prominent, and at the same time, they are facing increasingly complex security threats. Traditional static data classification and fixed access control strategies can no longer meet the security needs in a dynamic business environment. The leakage of sensitive enterprise data may not only cause huge economic losses, but also affect the reputation and competitiveness of the enterprise.
[0003] Existing data security access management methods often have the following problems: data sensitivity assessment lacks dynamism and context awareness, and cannot reflect changes in data value in a timely manner; access control policies are too rigid and difficult to adapt to complex and changing collaboration scenarios and user behavior patterns; encryption policies often adopt a one-size-fits-all approach, which not only affects data usage efficiency, but also fails to provide differentiated protection for data of different sensitivity levels. In addition, existing technologies also have shortcomings in abnormal behavior detection and threat response, making it difficult to effectively identify and respond to complex attacks such as advanced persistent threats. Summary of the invention
[0004] The present application provides a method and system for enterprise sensitive data security access management, which is used to achieve self-optimization of security policies and can continuously adapt to changing security environments and business needs.
[0005] In a first aspect, the present application provides a method for secure access management of enterprise sensitive data, the method comprising:
[0006] Conduct multi-dimensional analysis and dynamic evaluation of enterprise data to obtain classified labeled data with sensitivity index;
[0007] Performing association analysis based on the classification tag data and user attribute information to obtain an adaptive access control matrix;
[0008] According to the adaptive access control matrix, comprehensively record and analyze the user's access to the classified labeled data to obtain a data access audit log;
[0009] Based on the data access audit log, security domain division and collaboration strategy configuration are performed on the classified and marked data to obtain a virtual collaboration space and data sharing rules;
[0010] According to the sensitivity index, the virtual collaborative space and the data sharing rule, the classified marked data is differentially encrypted and a key is generated to obtain encrypted data and a decryption authority policy;
[0011] Based on the data access audit log, the decryption permission policy, the access record of the encrypted data and the external threat information, an integrated analysis is performed to obtain a dynamic defense strategy and a real-time response plan, and the adaptive access control matrix is updated according to the dynamic defense strategy to obtain a target access control matrix.
[0012] In a second aspect, the present application provides an enterprise sensitive data security access management system, the enterprise sensitive data security access management system comprising:
[0013] The evaluation module is used to conduct multi-dimensional analysis and dynamic evaluation of enterprise data to obtain classified labeled data with sensitivity index;
[0014] An association module, used for performing association analysis based on the classification mark data and user attribute information to obtain an adaptive access control matrix;
[0015] An analysis module, used to comprehensively record and analyze patterns of user access behaviors of the classified labeled data according to the adaptive access control matrix, and obtain a data access audit log;
[0016] A configuration module, used to perform security domain division and collaboration strategy configuration on the classified and marked data based on the data access audit log, to obtain a virtual collaboration space and data sharing rules;
[0017] A processing module, configured to perform differential encryption processing and key generation on the classified labeled data according to the sensitivity index, the virtual collaborative space and the data sharing rule, to obtain encrypted data and a decryption authority policy;
[0018] An update module is used to perform integrated analysis based on the data access audit log, the decryption permission policy, the access record of the encrypted data and the external threat information to obtain a dynamic defense strategy and a real-time response plan, and to update the adaptive access control matrix according to the dynamic defense strategy to obtain a target access control matrix.
[0019] The third aspect of the present application provides an enterprise sensitive data security access management device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the enterprise sensitive data security access management device executes the above-mentioned enterprise sensitive data security access management method.
[0020] A fourth aspect of the present application provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the above-mentioned enterprise sensitive data security access management method.
[0021] In the technical solution provided by this application, through multi-dimensional analysis and dynamic evaluation mechanism, the changes in data value and sensitivity can be reflected in real time. Based on the correlation analysis of user attributes and data characteristics, the dynamic adjustment of access control policies is realized, which not only ensures security but also improves the flexibility of data access. Through comprehensive recording and pattern analysis of user access behavior, potential abnormal behaviors and security risks can be discovered in a timely manner. Through security domain division and collaboration strategy configuration, the intelligent construction of virtual collaboration space is realized, which not only ensures data security but also promotes efficient collaboration across departments and organizations. According to data sensitivity and collaboration needs, differentiated encryption processing of different data is realized, which optimizes system performance while ensuring security. By integrating multi-source information for threat analysis, dynamic adjustment and real-time response of defense strategies are realized, and the system's ability to resist complex attacks is enhanced. By continuously updating the access control matrix, self-optimization of security policies is achieved, enabling the system to continuously adapt to changing security environments and business needs. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0023] Figure 1 This is a schematic diagram of an embodiment of a method for secure access management of sensitive enterprise data in an embodiment of the present application;
[0024] Figure 2 This is a schematic diagram of an embodiment of an enterprise sensitive data security access management system in an embodiment of the present application. DETAILED DESCRIPTION
[0025] Embodiments of the present application provide a method and system for secure access management of sensitive data of an enterprise. The terms "first", "second", "third", "fourth", etc. (if any) 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 precedence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof 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.
[0026] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the enterprise sensitive data security access management method in the embodiment of the present application includes:
[0027] Step S101: Perform multi-dimensional analysis and dynamic evaluation on enterprise data to obtain classified labeled data with sensitivity index;
[0028] It is understandable that the execution subject of the present application can be an enterprise sensitive data security access management system, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.
[0029] Specifically, the content of the enterprise data is parsed and features are extracted to obtain an initial feature vector set. The data topic distribution matrix is generated by clustering analysis and topic modeling of the initial feature vector set. By parsing the data content and extracting its feature vectors, similar data are grouped together using clustering technology, and the main topics and content distribution involved in the data are determined by topic modeling technology to form a topic distribution matrix that reflects the topic characteristics of the data. Semantic similarity calculation and association network construction are performed based on the generated data topic distribution matrix to obtain a data association map. The semantic similarity between each data item is calculated, and an association network is constructed based on the similarity results to form a data association map. The map identifies important nodes and their weights in the data through centrality analysis and importance evaluation, and generates a data node weight list. Centrality analysis helps determine the importance of data nodes in the network, while importance evaluation quantitatively scores the importance of these nodes. Based on the data node weight list and the predefined sensitive word library, sensitive information of the enterprise data is identified and quantitatively scored to obtain an initial sensitivity score. Sensitive information identification identifies sensitive information in data by matching words in the sensitive word library, and quantitative scoring assigns an initial sensitivity score to the data based on the identification results. Time decay modeling and dynamic weight calculation are performed on the initial sensitivity score to generate a sensitivity index. Time decay modeling takes into account the possibility that the sensitivity of data may change over time, while dynamic weight calculation ensures that the sensitivity index can be updated in real time. According to the generated sensitivity index, multi-level classification and label generation are performed on enterprise data to obtain initial classification labeling results. The data is divided into different levels through the sensitivity index and corresponding labels are generated, and the classification and labeling of the data are preliminarily completed. Based on the initial classification labeling results, classification integration and metadata encapsulation are performed to obtain classification labeling data with sensitivity index. The classification integration step integrates the relevant classification labels to ensure the accuracy and consistency of the classification results, while metadata encapsulation adds additional information to these classification labeling data for subsequent management and retrieval.
[0030] Step S102: performing association analysis based on the classification mark data and the user attribute information to obtain an adaptive access control matrix;
[0031] Specifically, the user attribute information is processed by feature vectorization, and the various attributes of the user are converted into quantifiable vector forms to obtain the user feature matrix. The user role is identified according to the user feature matrix, and the user's role in the system is identified by using the user attribute information to obtain the initial role division result. Based on the role division result and the classified labeled data, the access relationship between the user and the data is mined by association rules to obtain frequent item sets and association rule sets. By analyzing the user's access behavior to the data, it is identified which user roles frequently access which categories of data, and based on the frequent item sets and association rule sets, the user roles and data categories are constructed and weighted to form a role-data association network. Based on the constructed role-data association network, the user role centrality is calculated and the importance is sorted to obtain the role weight vector. The role's importance is determined by calculating its centrality in the network, and the association between the role and the data is quantified according to the role weight vectors and the sensitivity index of the classified labeled data to obtain the initial permission matrix. The initial permission matrix reflects the access rights level of each role to different categories of data. Based on the initial permission matrix, the permission allocation is analyzed for correlation, and the permission rule set is extracted from it. By analyzing the relationship between the permissions in the permission matrix, the implicit rules and patterns are identified, and the initial permission matrix is conditionally calculated and Bayesian network is constructed according to the permission rule set to generate a dynamic permission probability distribution. The dynamic permission probability distribution can dynamically reflect the access probability of each role to data under different conditions, helping to identify potential permission problems. Based on the dynamic permission probability distribution, permission allocation time series prediction and anomaly detection are performed to obtain a permission adjustment suggestion list. By analyzing the time series of permission allocation, the changes in future permission requirements are predicted, and anomalies in the current permission allocation are detected to generate adjustment suggestions. Based on the permission adjustment suggestion list, permission allocation multi-objective optimization and matrix reconstruction are performed to obtain an adaptive access control matrix. Multi-objective optimization takes into account the multiple objectives of permission allocation to ensure that permission allocation can meet security requirements and maximize the efficiency of system use. Matrix reconstruction adjusts the initial permission matrix based on the optimization results to form the final adaptive access control matrix.
[0032] Step S103: According to the adaptive access control matrix, the user's access to the classified marked data is comprehensively recorded and analyzed for patterns to obtain a data access audit log;
[0033] Specifically, the user access behavior is captured in real time and the event is serialized to generate the original access event stream. According to the original access event stream and the adaptive access control matrix, the user access behavior is verified and the compliance is checked to obtain the access compliance marking result. By capturing the user's access behavior in real time and converting these behaviors into a structured event sequence, it is helpful for subsequent permission verification and compliance checking. Based on the access compliance marking result, the access event is subjected to spatiotemporal feature extraction and behavior vectorization to obtain a multi-dimensional behavior feature matrix. Spatiotemporal feature extraction includes analyzing the time and spatial distribution characteristics of the access event, while behavior vectorization converts these characteristics into quantifiable vector form, thereby constructing a multi-dimensional behavior feature matrix. According to the multi-dimensional behavior feature matrix, the user access pattern is analyzed and the typical behavior is identified to obtain the behavior pattern prototype set. By identifying the user's typical access behavior pattern, the user's behavior can be better understood and predicted. Based on the behavior pattern prototype set and historical access data, behavior time series modeling and trend prediction are performed to generate a behavior evolution trajectory diagram. Time series modeling helps capture the law of user behavior changes over time, while trend prediction provides the possible evolution path of future behavior. According to the behavior evolution trajectory diagram and predefined anomaly indicators, the behavior deviation is calculated and the anomaly score is scored to obtain the behavior anomaly measurement table. By comparing the difference between the actual behavior and the predicted behavior, the behavior deviation is calculated and the anomaly score is assigned to form a behavior anomaly measurement table to reflect the degree of abnormality of the user behavior. Based on the behavior anomaly measurement table and the sensitivity index, the behavior risk level is evaluated and prioritized to obtain a risk event queue. The behavior risk level assessment evaluates the potential risks of user behavior based on the anomaly score and data sensitivity, and prioritizes these risks to form an orderly risk event queue. Through the risk event queue, audit task scheduling optimization and parallel processing are performed to generate a distributed audit task graph. Audit task scheduling optimization aims to improve the execution efficiency of audit tasks, while parallel processing ensures that multiple audit tasks can be performed simultaneously, thereby speeding up the audit process. Audit data incremental aggregation and compression encoding are performed based on the distributed audit task graph to generate structured audit log entries. Audit data incremental aggregation merges the newly generated audit data with the existing data, while compression encoding reduces the data storage space to form structured audit log entries. Audit information association analysis and causal reasoning are performed based on the structured audit log entries to obtain data access audit logs. Audit information correlation analysis identifies the relationships and patterns among different audit events by correlating them, while causal reasoning helps understand the causal relationship between events and generate data access audit logs.
[0034] Step S104: Based on the data access audit log, security domain division and collaboration strategy configuration are performed on the classified and marked data to obtain a virtual collaboration space and data sharing rules;
[0035] Specifically, the data access audit log is analyzed and the frequency is counted to generate a data access pattern matrix, which reflects the user's access frequency and behavior pattern to the data. Based on the data access pattern matrix and the classified labeled data, the data object similarity is calculated and grouped to obtain the initial data cluster set. Through similarity calculation, data objects with similar access patterns are grouped to form initial data clusters. Based on the initial data cluster set, the association relationship between data objects is modeled and community detection is performed to generate a data association community graph. Graph structure modeling represents data objects and their mutual associations through nodes and edges, and community detection identifies different communities or groups in the graph. Security level assessment and boundary division are performed based on the data association community graph to obtain preliminary security domain division results. Security level assessment determines the security level of each data community according to the importance and sensitivity of the data object and divides its boundaries. Based on the preliminary security domain division results, security domain overlapping area identification and conflict analysis are performed to obtain a security domain intersection matrix. Overlapping area identification and conflict analysis are aimed at discovering and resolving overlaps and conflicts between different security domains. Collaboration scenario modeling and simulation analysis are performed based on the security domain intersection matrix to generate a collaboration process map. Collaboration scenario modeling simulates the collaboration scenarios between different users and data objects, while simulation analysis helps optimize these scenarios to form a collaboration process map. Based on the collaboration process map and sensitivity index, data flow path risk assessment and control point setting are performed to obtain a data flow control strategy set. Risk assessment identifies potential risks in the data flow process and sets control measures at key points to form a data flow control strategy. Based on the data flow control strategy set, virtual collaboration environment topology design and resource allocation are performed to generate a virtual collaboration space. Topology design and resource allocation ensure that the virtual collaboration space can efficiently and securely support the collaboration needs of users. Based on the virtual collaboration space and user role permissions, data sharing behavior is extracted and formally described to obtain data sharing rules. By analyzing the behavior of users in the virtual collaboration space, effective sharing rules are extracted and formally described to ensure the clarity and enforceability of the rules.
[0036] Step S105: According to the sensitivity index, virtual collaborative space and data sharing rules, differentiated encryption processing and key generation are performed on the classified marked data to obtain encrypted data and decryption authority policy;
[0037] Specifically, the classified labeled data is processed in blocks and features are extracted to generate a data block feature vector set. By analyzing the feature vector set and sensitivity index, the encryption strength of the data block is evaluated and prioritized to form a hierarchical encryption task list. According to the sensitivity and characteristics of the data, appropriate encryption strength and priority are assigned to each data block. Based on the hierarchical encryption task list and virtual collaboration space, the appropriate encryption algorithm is automatically selected and parameter configuration is performed to form a differentiated encryption scheme. According to this encryption scheme, parallel encryption processing and integrity verification of the data block are performed to generate an initial encrypted data set. Parallel encryption processing uses multi-threading technology to speed up the encryption speed, while integrity verification ensures the integrity and correctness of the data during the encryption process. Based on the initial encrypted data set and data sharing rules, data hierarchical encapsulation and access control marking are performed to generate a multi-level protected encrypted data packet. Hierarchical encapsulation refers to stratifying data according to different security levels and setting corresponding access control markings for each layer of data, thereby achieving multi-level protection of the data. On this basis, fine-grained division of decryption permissions and policy generation are performed to obtain an initial decryption permission rule set. Fine-grained division ensures that each user can only access and decrypt data within his or her authority, thereby improving the security of data access. Based on the initial decryption permission rule set and the requirements of the collaborative scenario, the state machine modeling and transition rule definition of the decryption process are performed to form a dynamic decryption state diagram. State machine modeling is used to describe the various states and their transition rules in the decryption process, while the dynamic decryption state diagram shows these states and their transition paths. According to the dynamic decryption state diagram, risk assessment and threshold setting of the decryption operation are performed to obtain the decryption risk control parameters. Risk assessment helps identify the risks that may arise in the decryption process and control these risks by setting thresholds. Based on the decryption risk control parameters, key segmentation generation and distributed storage are performed to form a distributed key fragment set. Key segmentation generation divides the complete key into several fragments, while distributed storage stores these fragments in different secure locations, thereby improving the security of the key. According to the distributed key fragment set, the protocol design and security verification of the key reconstruction process are performed to ensure the security and reliability of key reconstruction, and finally generate encrypted data and decryption permission policies.
[0038] Step S106: Based on the data access audit log, decryption authority policy, encrypted data access record and external threat information, an integrated analysis is performed to obtain a dynamic defense strategy and a real-time response plan, and the adaptive access control matrix is updated according to the dynamic defense strategy to obtain a target access control matrix.
[0039] Specifically, the data access audit log and the access records of encrypted data are associated and pattern mined to generate a user behavior sequence diagram. By serializing the user's access behavior and combining it with the decryption permission policy, user operation compliance assessment and anomaly detection are performed to obtain a behavior anomaly measurement table. The behavior anomaly measurement table reflects whether the user behavior is normal or not, which helps to identify potential security threats. Security event correlation analysis is performed based on the behavior anomaly measurement table and external threat information to generate a threat event correlation network. By analyzing internal behavior anomalies and external threat information, related events are associated to identify potential threat sources and attack modes, and potential attack path reasoning and probability estimation are performed based on the threat event correlation network to obtain an attack path probability tree. The attack path probability tree is used to display various attack paths and their probability of occurrence, helping security personnel identify and prevent potential attacks. Defense resource optimization allocation and deployment planning are performed based on the attack path probability tree to generate a preliminary defense strategy plan. Defense resource optimization allocation achieves the maximum security protection effect at the lowest cost by reasonably allocating limited security resources. Cost-benefit evaluation and priority sorting of defense measures are performed based on the preliminary defense strategy plan to form a hierarchical defense strategy set. Cost-benefit evaluation takes into account the implementation cost and expected benefits of each defense measure, and prioritizes these measures to ensure that the most effective defense measures are implemented first. Defense rules are dynamically generated and conflict detection is performed based on the hierarchical defense strategy set to generate a target defense rule base. Dynamic generation of defense rules ensures that the defense strategy can respond to changing threats in a timely manner, while conflict detection avoids conflicts between different rules and ensures the effectiveness and consistency of defense rules. Emergency response processes are automatically orchestrated and decision trees are constructed based on the target defense rule base to generate real-time response plans. Automated orchestration of emergency response processes improves response speed and accuracy, while decision tree construction provides a clear emergency decision path to ensure that correct decisions can be made quickly in emergency situations. Access rights are dynamically adjusted and constraints are updated based on the real-time response plan to generate a temporary access control policy. The temporary access control policy dynamically adjusts the user's access rights according to the real-time response plan to ensure that the impact on normal business is minimized while responding to threats. According to the temporary access control policy, the adaptive access control matrix is globally optimized and balanced to generate a target access control matrix. Global optimization and balanced adjustment ensure that the access control matrix achieves the best balance between security and business continuity to form the final target access control matrix.
[0040] In the embodiments of the present application, through multi-dimensional analysis and dynamic evaluation mechanism, the changes in data value and sensitivity can be reflected in real time. Based on the correlation analysis of user attributes and data characteristics, the dynamic adjustment of access control policies is realized, which not only ensures security but also improves the flexibility of data access. Through comprehensive recording and pattern analysis of user access behavior, potential abnormal behaviors and security risks can be discovered in a timely manner. Through security domain division and collaboration strategy configuration, the intelligent construction of virtual collaboration space is realized, which not only ensures data security but also promotes efficient collaboration across departments and organizations. According to data sensitivity and collaboration needs, differentiated encryption processing of different data is realized, and system performance is optimized while ensuring security. By integrating multi-source information for threat analysis, dynamic adjustment and real-time response of defense strategies are realized, and the system's ability to resist complex attacks is enhanced. By continuously updating the access control matrix, self-optimization of security policies is achieved, enabling the system to continuously adapt to changing security environments and business needs.
[0041] In a specific embodiment, the process of executing step S101 may specifically include the following steps:
[0042] (1) Perform content analysis and feature extraction on enterprise data to obtain an initial feature vector set, and then perform cluster analysis and topic modeling on the initial feature vector set to obtain a data topic distribution matrix;
[0043] (2) Based on the data topic distribution matrix, semantic similarity calculation and association network construction are performed to obtain a data association graph. The data association graph is then subjected to centrality analysis and importance evaluation to obtain a data node weight list.
[0044] (3) Based on the data node weight list and the predefined sensitive word library, sensitive information of enterprise data is identified and quantitatively scored to obtain an initial sensitivity score. Based on the initial sensitivity score, time decay modeling and dynamic weight calculation are performed on the enterprise data to obtain a sensitivity index.
[0045] (4) Perform multi-level classification and label generation on enterprise data according to the sensitivity index to obtain initial classification and labeling results, and then perform classification integration and metadata encapsulation on the initial classification and labeling results to obtain classification and labeling data with sensitivity index.
[0046] Specifically, content analysis and feature extraction are performed on enterprise data to obtain an initial feature vector set. Content analysis refers to the structured processing of raw data to extract key information, such as text, numerical values, images, etc. Through natural language processing technology, text data is converted into vector representations, which capture the semantic and structural information of the data. For example, each document is represented as a vector using technologies such as TF-IDF (term frequency-inverse document frequency) or Word2Vec. Cluster analysis and topic modeling are performed on the initial feature vector set to generate a data topic distribution matrix. Cluster analysis (such as K-means or DBSCAN) groups similar data items into one category, while topic modeling (such as LDA-Latent Dirichlet Allocation) identifies hidden topics in the data. For example, suppose represents the initial feature vector set, ,in It is The feature vector of each data item. Through LDA, we get the data topic distribution matrix , ,in Indicates The distribution of topics. Based on the data topic distribution matrix, semantic similarity calculation and association network construction are performed to obtain a data association graph. Semantic similarity calculation can use methods such as cosine similarity, Euclidean distance or Jaccard coefficient. For example, for two feature vectors and , the cosine similarity It can be expressed as:
[0047] ;
[0048] Through similarity calculation, a data association network is constructed, where nodes represent data items and edges represent similarity relationships between data items. The data association graph is analyzed for centrality and importance, and a list of data node weights is generated. Centrality analysis methods include degree centrality, betweenness centrality, and eigenvector centrality. For example, a node The degree centrality It is expressed as:
[0049] ;
[0050] in is an element of the adjacency matrix, representing the node and The importance of each data node is determined through centrality analysis and importance assessment, thereby generating a data node weight list. Based on the data node weight list and the predefined sensitive word library, sensitive information is identified and quantitatively scored for enterprise data to generate an initial sensitivity score. Sensitive information identification detects potential sensitive content by matching keywords or patterns in the data. Quantitative scoring assigns an initial sensitivity score to each data item based on the amount and type of sensitive information. For example, Indicates The weight of the data node, Indicates The number of sensitive information in the data items, then the initial sensitivity score It can be expressed as:
[0051] ;
[0052] in is a function that maps weights and the amount of sensitive information to a sensitivity score. Based on the initial sensitivity score, time decay modeling and dynamic weight calculation are performed on enterprise data to generate a sensitivity index. Time decay modeling considers the change of data sensitivity over time and usually uses an exponential decay model. For example, Represents a data item In time The sensitivity score of represents the attenuation coefficient, then the sensitivity index It can be expressed as:
[0053] ;
[0054] in, It reflects the sensitivity of the data item at the current time. According to the sensitivity index, enterprise data is classified and labeled at multiple levels to obtain the initial classification labeling results. Data classification can divide data into multiple levels according to the sensitivity index, such as high sensitivity, medium sensitivity and low sensitivity. Data items at each level will be marked with corresponding labels for subsequent management and protection. Based on the initial classification labeling results, classification integration and metadata encapsulation are performed to generate classification labeled data with sensitivity index. Classification integration refers to the integration of data items of the same or similar categories to form a unified classification result, while metadata encapsulation adds additional information to each classification labeled data, such as creation time, creator and data source, for subsequent retrieval and management.
[0055] In a specific embodiment, the process of executing step S102 may specifically include the following steps:
[0056] (1) Perform feature vectorization on user attribute information to obtain a user feature matrix, and perform user role identification based on the user feature matrix to obtain the initial role division result;
[0057] (2) Based on the initial role division results and classified labeled data, association rule mining is performed on the user-data access relationship to obtain frequent item sets and association rule sets. Based on the frequent item sets and association rule sets, a bipartite graph is constructed and weights are assigned for user roles and data categories to obtain a role-data association network.
[0058] (3) Based on the role-data association network, the user role centrality is calculated and the importance is ranked to obtain the role weight vector. Based on the role weight vector and the sensitivity index, the permission level of the role-data association is quantified to obtain the initial permission matrix.
[0059] (4) Based on the initial permission matrix, the permission allocation is analyzed for correlation to obtain the permission rule set. Based on the permission rule set, the initial permission matrix is subjected to conditional probability calculation and Bayesian network construction to obtain the dynamic permission probability distribution.
[0060] (5) Based on the dynamic permission probability distribution, permission allocation time series prediction and anomaly detection are performed to obtain a permission adjustment suggestion list. Based on the permission adjustment suggestion list, permission allocation multi-objective optimization and matrix reconstruction are performed to obtain an adaptive access control matrix.
[0061] Specifically, the user attribute information is first processed by feature vectorization to generate a user feature matrix. The various attributes of the user (such as department, position, experience, authority level, etc.) are converted into vector representations for subsequent calculations and analysis. For example, if the user set is , the attribute set of each user is , then the user feature matrix It can be expressed as:
[0062] ;
[0063] in, Indicates user No. Attributes. User role identification is performed based on the user feature matrix to obtain the initial role division result. Role identification can divide users into different role groups through clustering algorithms (such as K-means or hierarchical clustering). For example, using the K-means clustering algorithm, the user feature vector set Divide into role groups, each of which represents an initial role division result. Based on the initial role division results and the classified labeled data, association rule mining is performed on the user-data access relationship to obtain frequent item sets and association rule sets. Association rule mining is to identify common patterns between user roles and data categories by analyzing the user's data access behavior. For example, using the Apriori algorithm, frequent item sets represent combinations of data items that often appear together, and association rules represent the access patterns of these combinations. Assume is the initial character set, is a data category set, then the frequent item set and association rule set can be expressed as , indicating the role Frequently accessed data categories According to the frequent item sets and association rule sets, a bipartite graph is constructed and weights are assigned to user roles and data categories to obtain a role-data association network. A bipartite graph is a graph composed of two different types of nodes (user roles and data categories), where the edge represents the access relationship between the role and the data category, and the weight of the edge represents the access frequency. For example, suppose is the role-data association network, is its adjacency matrix, then Representing roles With data category The access relationship weight between them. Based on the role-data association network, the user role centrality is calculated and the importance is sorted to obtain the role weight vector. The centrality calculation is to evaluate the importance of each role in the association network. Commonly used methods include degree centrality, betweenness centrality and eigenvector centrality. Representing roles The degree centrality of is calculated as: ;
[0064] By calculating the centrality of each role, we get the role weight vector ,in Representing roles According to the role weight vector and sensitivity index, the permission level of the role-data association is quantified to generate the initial permission matrix. Permission level quantification is to assign appropriate access rights based on the importance of the role and the sensitivity of the data. For example, is the initial authority matrix, is the sensitivity index matrix, then Representing roles For data categories The permission level can be calculated by the following formula: ;
[0065] in, Indicates data category The sensitivity index of the permission matrix is calculated. Based on the initial permission matrix, the permission allocation is analyzed for correlation and the permission rule set is obtained. Correlation analysis is to identify implicit rules and patterns by analyzing the relationship between the permissions in the permission matrix. According to the permission rule set, the initial permission matrix is calculated for conditional probability and Bayesian network is constructed to obtain the dynamic permission probability distribution. Bayesian network is a probabilistic graph model that describes the relationship between variables through conditional probability. Assume is a Bayesian network, then Representing roles For data categories The access probability of the permission allocation is . Based on the dynamic permission probability distribution, permission allocation time series prediction and anomaly detection are performed to obtain a permission adjustment suggestion list. Time series prediction is to predict future permission requirements by analyzing historical data of permission allocation, and anomaly detection is to identify anomalies in permission allocation. Through time series prediction and anomaly detection, a permission adjustment suggestion list is generated to indicate which permissions need to be adjusted. Based on the permission adjustment suggestion list, permission allocation multi-objective optimization and matrix reconstruction are performed to generate an adaptive access control matrix. Multi-objective optimization optimizes the permission allocation scheme by balancing security and efficiency, and matrix reconstruction adjusts the initial permission matrix according to the optimization results to form the final adaptive access control matrix. For example, suppose is the initial authority matrix, is the optimization matrix, then the adaptive access control matrix It can be expressed as: in, To adjust the coefficient, control the influence of the optimization matrix on the authority allocation.
[0066] In a specific embodiment, the process of executing step S103 may specifically include the following steps:
[0067] (1) Real-time capture and event serialization of user access behavior to obtain the original access event stream. Based on the original access event stream and the adaptive access control matrix, the user access behavior is verified for authority and compliance, and the access compliance marking result is obtained.
[0068] (2) Based on the access compliance tagging results, the access events are subjected to spatiotemporal feature extraction and behavior vectorization to obtain a multi-dimensional behavior feature matrix. Based on the multi-dimensional behavior feature matrix, the user access patterns are analyzed and typical behaviors are identified to obtain a behavior pattern prototype set.
[0069] (3) Based on the behavioral pattern prototype set and historical access data, behavioral time series modeling and trend prediction are performed to obtain a behavioral evolution trajectory diagram. Behavioral deviation calculation and anomaly scoring are performed based on the behavioral evolution trajectory diagram and predefined anomaly indicators to obtain a behavioral anomaly scale;
[0070] (4) Based on the behavior anomaly scale and sensitivity index, the behavior risk level is evaluated and prioritized to obtain a risk event queue. The audit task scheduling optimization and parallel processing are performed through the risk event queue to obtain a distributed audit task graph.
[0071] (5) Based on the distributed audit task graph, audit data is incrementally aggregated and compressed to obtain structured audit log entries. Audit information association analysis and causal reasoning are performed based on the structured audit log entries to obtain data access audit logs.
[0072] Specifically, the user access behavior is captured in real time and the event is serialized to generate the original access event stream. Real-time capture means that when the user accesses the system, his operation behavior is recorded immediately, including the accessed resources, timestamps, operation types, etc. Event serialization converts these access behaviors into an ordered event stream for subsequent processing. Let the event stream be , where each event Include access time , User ID , Resource Identifier and operation type According to the original access event flow and the adaptive access control matrix, the user's access behavior is verified and compliance checked to obtain the access compliance tag result. The permission verification checks whether the user has the corresponding permissions by comparing the user's access behavior with the access control matrix; the compliance check evaluates whether the user's behavior complies with the established security policies and regulatory requirements. For example, let the adaptive access control matrix be , then for the event , the result of its permission verification It can be expressed as:
[0073] ;
[0074] like , it means that the user About Resources The access is legal, otherwise it is illegal. The compliance check result is combined with the permission verification result to form the access compliance marking result. Based on the access compliance tagging results, the access events are subjected to spatiotemporal feature extraction and behavior vectorization to generate a multi-dimensional behavior feature matrix. Spatiotemporal feature extraction is to extract temporal and spatial features from access events, such as access frequency, access time interval, geographic location, etc. Behavior vectorization is to convert these features into vector form to represent the user's access behavior. is a multi-dimensional behavior feature matrix, where each row represents a user's access behavior vector ,but It can be expressed as:
[0075]
[0076] in, Indicates user The behavioral feature vector, It is eigenvalues, such as the number of visits, the duration of visits, etc. According to the multi-dimensional behavior feature matrix, the user access pattern is analyzed and typical behaviors are identified to obtain a set of behavior pattern prototypes. Access pattern analysis uses clustering algorithms (such as K-means or DBSCAN) to cluster similar behavior feature vectors and identify typical access patterns. For example, using the K-means clustering algorithm, the behavior feature matrix Divide into Behavior Prototype ,in , each Represents a typical behavior pattern. Based on the behavior pattern prototype set and historical access data, behavior time series modeling and trend prediction are performed to generate a behavior evolution trajectory diagram. Time series modeling predicts future behavior trends by analyzing the time series of user behavior. For example, using the ARIMA model (autoregressive integrated moving average model) for time series prediction, assume that the user In time The behavioral characteristics of , then its future behavior characteristics It can be expressed as: ;
[0077] in, is the autoregression coefficient, is the sliding average coefficient, is a white noise term. Through time series modeling and trend prediction, the evolution trajectory of user behavior is generated. Behavior deviation calculation and anomaly scoring are performed based on the behavior evolution trajectory and predefined anomaly indicators to generate a behavior anomaly measurement table. Behavior deviation calculation is to calculate the degree of anomaly by comparing the difference between actual behavior and predicted behavior. For users In time The behavioral deviation vector of The deviation of a feature can be expressed as: ;
[0078] in, It is the predicted The anomaly score is a comprehensive assessment of behavioral deviations to obtain an anomaly measurement value. , reflecting the user's time Based on the behavior anomaly scale and sensitivity index, the behavior risk level is evaluated and prioritized to obtain the risk event queue. The behavior risk level assessment is to evaluate the risk level of user behavior by combining the anomaly score and data sensitivity index. In time The risk level is , then it can be expressed as:
[0079] ;
[0080] in, and is the weight coefficient, is the data sensitivity index. Priority sorting is to sort risk events according to risk level to form a risk event queue. Audit task scheduling optimization and parallel processing are performed through the risk event queue to generate a distributed audit task graph. Audit task scheduling optimization is to improve audit efficiency by reasonably allocating audit resources, and parallel processing is to speed up audit by executing multiple audit tasks in parallel. The distributed audit task graph represents each audit task and its dependencies. Audit data incremental aggregation and compression coding are performed based on the distributed audit task graph to generate structured audit log entries. Audit data incremental aggregation merges newly generated audit data with existing data, and compression coding compresses audit data to reduce storage space. Structured audit log entries contain information such as timestamp, user ID, operation type, resource ID, etc., which is convenient for subsequent analysis. Audit information association analysis and causal reasoning are performed based on structured audit log entries to generate data access audit logs. Audit information association analysis is to identify potential security threats and abnormal behaviors by analyzing the association between different audit events. Causal reasoning is to determine the cause and impact of an event by analyzing the causal relationship between events. Data access audit logs are a comprehensive record of all audit events, reflecting the security status and user behavior of the system.
[0081] In a specific embodiment, the process of executing step S104 may specifically include the following steps:
[0082] (1) Analyze and perform frequency statistics on data access audit logs to obtain a data access pattern matrix. Then, calculate and group data objects based on the data access pattern matrix and classified labeled data to obtain an initial data cluster set.
[0083] (2) Based on the initial data cluster set, the relationship between data objects is modeled and community detected to obtain a data association community graph. Security level assessment and boundary division are performed based on the data association community graph to obtain preliminary security domain division results.
[0084] (3) Based on the preliminary security domain division results, the overlapping areas of security domains are identified and conflicts are analyzed to obtain the security domain intersection matrix. The collaboration scenario modeling and simulation analysis are then performed based on the security domain intersection matrix to obtain the collaboration process map.
[0085] (4) Based on the collaborative process map and sensitivity index, data flow path risk assessment and control point setting are performed to obtain a data flow control strategy set. Based on the data flow control strategy set, virtual collaborative environment topology design and resource allocation are performed to obtain a virtual collaborative space.
[0086] (5) Based on the virtual collaborative space and user role permissions, the data sharing behavior is extracted and formally described to obtain data sharing rules.
[0087] Specifically, the data access audit log is analyzed and frequency counted to generate a data access pattern matrix. The data access audit log contains all the user access records to the data in the system, including the access time, access type, user identity and the data object accessed. By performing frequency statistics on the log, the frequency of access to each data object is determined, and a data access pattern matrix is constructed. . Let the data object set be , the user set is , then the data access pattern matrix Elements Indicates user Accessing Data Objects The frequency of the matrix The form is as follows:
[0088] ;
[0089] According to the data access pattern matrix and the classified labeled data, the data objects are similarly calculated and grouped to obtain the initial data cluster set. The similarity calculation can be done using cosine similarity. Suppose the data objects The access vector is and , then the similarity It can be expressed as: ;
[0090] Through similarity calculation, data objects with high similarity are grouped together to form an initial data cluster set. Based on the initial data cluster set, graph structure modeling and community detection are performed on the association relationship between data objects to obtain a data association community graph. Graph structure modeling represents data objects as nodes of a graph, and establishes edges between data objects with high similarity to construct a data association graph. Community detection clusters data objects with strong associations by identifying the community structure in the graph. Commonly used community detection algorithms include the Louvain algorithm and the Girvan-Newman algorithm, which can be used to obtain a data association community graph. Security level assessment and boundary division are performed based on the data association community graph to obtain preliminary security domain division results. Security level assessment is to assess the security level of each community based on the sensitivity and importance of the data object, and to divide the boundaries of the data object based on the assessment results to form preliminary security domain division results. Based on the preliminary security domain division results, security domain overlapping area identification and conflict analysis are performed to obtain a security domain intersection matrix. Security domain overlapping area identification is to identify possible conflict points by analyzing the overlapping areas between different security domains, and to perform detailed conflict analysis. Let the security domain be , then the security domain cross matrix Elements Indicates security domain and The overlap between the security domains can be intuitively understood through the cross matrix. Collaboration scenario modeling and simulation analysis are performed according to the security domain cross matrix to obtain a collaborative process map. Collaboration scenario modeling is to simulate the data flow and collaboration process between different security domains, establish a collaborative scenario model, and perform simulation analysis to optimize the collaborative process. The collaborative process map is a graphical representation of the collaborative process, showing the data flow and collaboration relationship between different security domains. Data flow path risk assessment and control point setting are performed based on the collaborative process map and sensitivity index to obtain a data flow control policy set. Data flow path risk assessment is to analyze the path of data flow between different security domains, identify potential risk points, and set control measures at these points to reduce risks. Control point setting is to determine the key control points of data flow based on the risk assessment results and formulate corresponding control policies. Virtual collaborative environment topology design and resource allocation are performed based on the data flow control policy set to obtain a virtual collaborative space. Virtual collaborative environment topology design is to ensure the secure flow of data between different security domains by designing the network topology of the virtual collaborative environment. Resource allocation is to reasonably allocate computing resources and storage resources based on the topology design results to support the collaborative process. Based on the virtual collaborative space and user role permissions, the data sharing behavior is extracted and formally described to obtain data sharing rules. Rule extraction is to extract effective sharing rules by analyzing the data sharing behavior of users in the virtual collaborative space, and then formally describe them to ensure the clarity and enforceability of the rules.
[0091] In a specific embodiment, the process of executing step S105 may specifically include the following steps:
[0092] (1) Block processing and feature extraction are performed on the classified labeled data to obtain a data block feature vector set. The encryption strength of the data blocks is evaluated and prioritized based on the data block feature vector set and the sensitivity index to obtain a hierarchical encryption task list.
[0093] (2) Based on the hierarchical encryption task list and virtual collaboration space, the encryption algorithm is automatically selected and parameter configuration is performed to obtain a differentiated encryption scheme. The data blocks are then encrypted in parallel and integrity checked according to the differentiated encryption scheme to obtain the initial encrypted data set.
[0094] (3) Based on the initial encrypted data set and data sharing rules, data is hierarchically encapsulated and access control marked to obtain a multi-level protected encrypted data package. Decryption permissions are fine-grainedly divided and policies are generated based on the multi-level protected encrypted data package to obtain an initial decryption permission rule set;
[0095] (4) Based on the initial decryption authority rule set and the requirements of the collaborative scenario, the decryption process state machine is modeled and the transition rules are defined to obtain a dynamic decryption state diagram. The decryption operation risk assessment and threshold setting are performed based on the dynamic decryption state diagram to obtain the decryption risk control parameters;
[0096] (5) Based on the decryption risk control parameters, the key is segmented and distributedly stored to obtain a distributed key fragment set. The protocol design and security verification of the key reconstruction process are performed based on the distributed key fragment set to obtain the encrypted data and decryption permission policy.
[0097] Specifically, the entire data set is processed in blocks, and a large data set is decomposed into several smaller data blocks for feature extraction and subsequent encryption processing. For each data block, its feature vector set is extracted. Feature extraction includes extracting content features, structural features, and access features of the data block. Suppose each data block is , whose feature vector set is ,in Indicates The feature vector of each data block. The encryption strength of the data blocks is evaluated and prioritized according to the feature vector set and sensitivity index of the data blocks. The encryption strength evaluation is to determine the encryption strength required for each data block according to the characteristics and sensitivity index of the data block. The sensitivity index is , the encryption strength evaluation function is , then The encryption strength of a data block can be expressed as: ;
[0098] in, and is the weight coefficient, The norm of the feature vector of the data block. According to the encryption strength evaluation results, the data blocks are prioritized to form a hierarchical encryption task list. Based on the hierarchical encryption task list and the virtual collaboration space, the encryption algorithm is automatically selected and parameter configuration is performed to generate a differentiated encryption scheme. The selection of the encryption algorithm is based on the encryption strength of the data block and the requirements of the virtual collaboration space. Common encryption algorithms include AES, RSA, and ECC. The parameter configuration sets the corresponding key length, encryption mode, etc. according to the different algorithm selections. The differentiated encryption scheme ensures that each data block uses the appropriate encryption algorithm and parameters to improve the security and efficiency of encryption. According to the differentiated encryption scheme, the data blocks are encrypted in parallel and integrity checked to generate an initial encrypted data set. Parallel encryption processing uses multi-threaded or distributed computing technology to encrypt multiple data blocks at the same time to improve encryption efficiency. Integrity checking ensures that the data has not been tampered with during the encryption process by calculating the checksum (such as SHA-256) of each data block. The initial encrypted data set contains all encrypted data blocks and their checksums. Data hierarchical encapsulation and access control marking are performed based on the initial encrypted data set and data sharing rules to generate a multi-level protected encrypted data packet. The encrypted data is layered according to different security levels, and each layer of the data packet contains the corresponding access control tags to ensure that only users with corresponding permissions can access data at a specific level. Access control tags are generated according to data sharing rules, which stipulate the access rights and decryption conditions of each user. Fine-grained division of decryption permissions and policy generation are performed based on the encrypted data packets with multi-level protection to form an initial decryption permission rule set. Fine-grained division is to refine the decryption permissions to each data block to ensure precise control of the decryption process. Policy generation defines decryption rules based on the user's role and permissions, including the conditions and restrictions for the use of decryption keys. Based on the initial decryption permission rule set and collaborative scenario requirements, state machine modeling and transition rule definition of the decryption process are performed to generate a dynamic decryption state diagram. State machine modeling is to describe the dynamic behavior of the decryption operation by defining each state in the decryption process and its transition conditions. Let the state set be , the transformation rule set is , then the dynamic decryption state graph can be represented as a directed graph, where nodes represent states and edges represent state transition rules. According to the dynamic decryption state graph, risk assessment and threshold setting of decryption operation are performed to obtain decryption risk control parameters. Risk assessment is to analyze the risk points that may appear in the decryption process and set corresponding control thresholds to reduce the risk. Let the risk assessment function be , indicating that in the state To status During the conversion process, the risk value is:
[0099] ;
[0100] in, is the risk adjustment factor, Indicates from the state To status The conversion probability, Represents the potential impact of the conversion. Based on the decryption risk control parameters, key segmentation generation and distributed storage are performed to obtain a distributed key fragment set. Key segmentation generation is to decompose the complete key into several fragments, each fragment is stored in a different secure location to ensure the security of the key. Distributed storage uses a distributed storage system to store key fragments in a decentralized manner to improve security and fault tolerance. According to the distributed key fragment set, the protocol design and security verification of the key reconstruction process are performed to generate encrypted data and decryption permission policies. The key reconstruction process is to recombine the decentralized stored key fragments through a predefined protocol to generate a complete key for decryption operations. Security verification ensures the security of the key reconstruction process by checking the integrity and correctness of the key fragments.
[0101] In a specific embodiment, the process of executing step S106 may specifically include the following steps:
[0102] (1) Perform association and pattern mining on data access audit logs and encrypted data access records to obtain a user behavior sequence diagram. Perform user operation compliance assessment and anomaly detection based on the user behavior sequence diagram and decryption permission policy to obtain a behavior anomaly measurement table.
[0103] (2) Perform security event correlation analysis based on the behavior anomaly measurement table and external threat information to obtain the threat event correlation network, and perform potential attack path reasoning and probability estimation based on the threat event correlation network to obtain the attack path probability tree;
[0104] (3) Based on the attack path probability tree, defense resources are optimally allocated and deployed to obtain a preliminary defense strategy plan. Based on the preliminary defense strategy plan, the cost-benefit evaluation and priority ranking of defense measures are performed to obtain a hierarchical defense strategy set.
[0105] (4) Dynamically generate defense rules and detect conflicts based on the hierarchical defense strategy set to obtain a target defense rule base. Automatically orchestrate the emergency response process and build a decision tree based on the target defense rule base to obtain a real-time response plan.
[0106] (5) Based on the real-time response scheme, the access rights are dynamically adjusted and the constraints are updated to obtain a temporary access control strategy. Based on the temporary access control strategy, the adaptive access control matrix is globally optimized and balanced to obtain the target access control matrix.
[0107] Specifically, the data access audit log and the access records of encrypted data are associated and pattern mined to generate a user behavior sequence diagram. The data access audit log contains all the user's operation records on the system, while the access records of encrypted data contain the user's decryption and use of encrypted data. By associating these records, the user's behavior pattern is identified and represented as a sequence diagram, where the nodes of the sequence diagram represent user operations and the edges represent the order and association of operations. According to the user behavior sequence diagram and the decryption permission policy, user operation compliance assessment and anomaly detection are performed to generate a behavior anomaly measurement table. Compliance assessment is to check the legality of the operation by comparing the user's actual operation with the decryption permission policy. Anomaly detection is to mark potential security threats by identifying operations that deviate from normal behavior patterns. Let the user operation set be , the decryption permission policy is , then the user The operational compliance can be expressed as: ;
[0108] in, Is an indicator function, indicating the operation Whether it complies with the decryption permission policy. Based on the compliance assessment results and anomaly detection, a behavior anomaly measurement table is generated to reflect the degree of abnormal behavior of each user. Security event correlation analysis is performed based on the behavior anomaly measurement table and external threat information to obtain a threat event correlation network. External threat information includes external security threats such as network attacks and malware. By correlating and analyzing the behavior anomaly measurement table with external threat information, potential threat events are identified and a threat event correlation network is constructed. Suppose the threat event set is , then the threat event association network is , where the nodes of the threat event association network represent threat events, and the edges represent the associations between events. Based on the threat event association network, potential attack paths are inferred and probabilities are estimated to obtain an attack path probability tree. Attack path inference is to identify possible attack paths and estimate their occurrence probabilities by analyzing the associations between threat events. Suppose the attack path is , the probability of its occurrence is , then the attack path probability tree can be expressed as:
[0109] ;
[0110] in, For events The probability of occurrence of each attack path is calculated, the most likely attack path is identified, and defense resources are optimally allocated and deployed. Defense resources are optimally allocated and deployed based on the attack path probability tree to generate a preliminary defense strategy. Defense resource optimization is to achieve the maximum defense effect at the lowest cost by reasonably allocating limited defense resources. Deployment planning is to formulate a specific defense measure deployment plan based on the allocation results of defense resources. Cost-benefit evaluation and priority sorting of defense measures are performed according to the preliminary defense strategy to generate a hierarchical defense strategy set. Cost-benefit evaluation is to determine the priority of each defense measure by comparing its implementation cost and expected benefit, and to form an ordered defense strategy set. Defense rules are dynamically generated and conflict detected based on the hierarchical defense strategy set to generate a target defense rule base. Defense rule dynamic generation is to ensure that it can respond to changing security threats by updating defense rules in real time. Conflict detection is to identify and resolve conflicts between different defense rules to ensure the effectiveness and consistency of defense rules. The target defense rule base is a collection of all valid defense rules, which is used to guide actual defense measures. According to the target defense rule base, the emergency response process is automatically orchestrated and the decision tree is constructed to obtain a real-time response plan. The automated orchestration of emergency response processes is to orchestrate predefined emergency response processes through automated tools to improve response speed and accuracy. The decision tree construction is to form a clear decision tree by analyzing the decision paths of different emergency response processes to guide the actual emergency response operations. The real-time response plan is a set of all emergency response processes and decision paths, which is used to guide the actual emergency response operations. Based on the real-time response plan, the access rights are dynamically adjusted and the constraints are updated to obtain a temporary access control policy. The dynamic adjustment of access rights is to dynamically adjust the user's access rights according to the real-time response plan to respond to current security threats. The constraint update is to ensure the effectiveness and flexibility of the access control policy by updating the constraints of the access control policy in real time. The temporary access control policy is a set of current access control policies and constraints, which is used to guide the actual access control operations. According to the temporary access control policy, the adaptive access control matrix is globally optimized and balanced to obtain the target access control matrix. Global optimization is to ensure the security and availability of the system by adjusting the access control policy as a whole. Balanced adjustment is to ensure the fairness and rationality of the access control policy by balancing the access requirements between different users and resources. The target access control matrix is a collection of all optimization and adjustment results, which is used to guide the actual access control operations.
[0111] The above describes the enterprise sensitive data security access management method in the embodiment of the present application. The following describes the enterprise sensitive data security access management system in the embodiment of the present application. Figure 2In the embodiment of the present application, an embodiment of the enterprise sensitive data security access management system includes:
[0112] Evaluation module 201, used to perform multi-dimensional analysis and dynamic evaluation on enterprise data to obtain classified labeled data with sensitivity index;
[0113] An association module 202, used to perform association analysis based on the classification tag data and the user attribute information to obtain an adaptive access control matrix;
[0114] The analysis module 203 is used to comprehensively record and analyze the user's access to the classified and marked data according to the adaptive access control matrix to obtain a data access audit log;
[0115] Configuration module 204, used to divide the classified and marked data into security domains and configure collaboration policies based on the data access audit log, and obtain virtual collaboration space and data sharing rules;
[0116] The processing module 205 is used to perform differential encryption processing and key generation on the classified marked data according to the sensitivity index, virtual collaborative space and data sharing rules to obtain encrypted data and decryption authority policy;
[0117] The update module 206 is used to perform integrated analysis based on data access audit logs, decryption permission policies, encrypted data access records and external threat information to obtain dynamic defense strategies and real-time response plans, and update the adaptive access control matrix according to the dynamic defense strategy to obtain the target access control matrix.
[0118] Through the collaboration of the above components, multi-dimensional analysis and dynamic evaluation mechanism, the changes in data value and sensitivity can be reflected in real time. Based on the correlation analysis of user attributes and data characteristics, the dynamic adjustment of access control policies is realized, which not only ensures security but also improves the flexibility of data access. Through comprehensive recording and pattern analysis of user access behavior, potential abnormal behaviors and security risks can be discovered in a timely manner. Through security domain division and collaboration strategy configuration, the intelligent construction of virtual collaboration space is realized, which not only ensures data security but also promotes efficient collaboration across departments and organizations. According to data sensitivity and collaboration needs, differentiated encryption processing of different data is realized, which optimizes system performance while ensuring security. By integrating multi-source information for threat analysis, dynamic adjustment and real-time response of defense strategies are realized, and the system's ability to resist complex attacks is enhanced. By continuously updating the access control matrix, self-optimization of security strategies is achieved, enabling the system to continuously adapt to changing security environments and business needs.
[0119] The present application also provides an enterprise sensitive data security access management device, which includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the enterprise sensitive data security access management method in the above-mentioned embodiments.
[0120] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the enterprise sensitive data security access management method.
[0121] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0122] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.
[0123] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for secure access management of sensitive enterprise data, characterized in that: The method comprises: Conduct multi-dimensional analysis and dynamic evaluation of enterprise data to obtain classified labeled data with sensitivity index; Performing association analysis based on the classification tag data and user attribute information to obtain an adaptive access control matrix; According to the adaptive access control matrix, comprehensively record and analyze the user's access to the classified labeled data to obtain a data access audit log; Based on the data access audit log, the classified labeled data is divided into security domains and collaborative strategy configurations are performed to obtain a virtual collaborative space and data sharing rules; specifically, the data access audit log is analyzed and frequency statistics are performed to obtain a data access pattern matrix, and data object similarity is calculated and grouped according to the data access pattern matrix and the classified labeled data to obtain an initial data cluster set; based on the initial data cluster set, the association relationship between data objects is modeled and community detected to obtain a data association community graph, and security level assessment and boundary division are performed based on the data association community graph to obtain a preliminary security Domain division results; based on the preliminary security domain division results, security domain overlapping area identification and conflict analysis are performed to obtain a security domain intersection matrix, and collaborative scenario modeling and simulation analysis are performed based on the security domain intersection matrix to obtain a collaborative process map; based on the collaborative process map and the sensitivity index, data flow path risk assessment and control point setting are performed to obtain a data flow control strategy set, and based on the data flow control strategy set, virtual collaborative environment topology design and resource allocation are performed to obtain a virtual collaborative space; based on the virtual collaborative space and user role permissions, data sharing behavior is subjected to rule extraction and formal description to obtain data sharing rules; According to the sensitivity index, the virtual collaboration space and the data sharing rules, the classified marked data is differentially encrypted and keys are generated to obtain encrypted data and decryption permission policies; specifically including: block processing and feature extraction of the classified marked data to obtain a data block feature vector set, and data block encryption strength evaluation and priority sorting are performed according to the data block feature vector set and the sensitivity index to obtain a hierarchical encryption task list; based on the hierarchical encryption task list and the virtual collaboration space, encryption algorithm is automatically selected and parameter configuration is performed to obtain a differentiated encryption scheme, and data block parallel encryption processing and integrity verification are performed according to the differentiated encryption scheme to obtain an initial encrypted data set; based on the initial encrypted data set and the data sharing rules, data layered encapsulation and access control marking are performed to obtain an encrypted data packet with multi-level protection, and according to the multi-level protection encryption The data packet is divided into fine-grained decryption permissions and policies are generated to obtain an initial decryption permission rule set; based on the initial decryption permission rule set and collaborative scenario requirements, state machine modeling and conversion rule definition of the decryption process are performed to form a dynamic decryption state diagram. State machine modeling describes the dynamic behavior of the decryption operation by defining each state and its conversion conditions in the decryption process, while the dynamic decryption state diagram shows these states and their conversion paths. The dynamic decryption state diagram is represented as a directed graph, in which nodes represent states and edges represent state conversion rules. Decryption operation risk assessment and threshold setting are performed based on the dynamic decryption state diagram to obtain decryption risk control parameters; based on the decryption risk control parameters, key segmentation generation and distributed storage are performed to obtain a distributed key fragment set, and protocol design and security verification of the key reconstruction process are performed based on the distributed key fragment set to obtain encrypted data and decryption permission policies; Based on the data access audit log, the decryption permission policy, the access record of the encrypted data and the external threat information, an integrated analysis is performed to obtain a dynamic defense strategy and a real-time response plan, and the adaptive access control matrix is updated according to the dynamic defense strategy to obtain a target access control matrix.
2. The enterprise sensitive data security access management method according to claim 1 is characterized in that: The multi-dimensional analysis and dynamic evaluation of enterprise data to obtain classified labeled data with sensitivity indexes include: Performing content analysis and feature extraction on enterprise data to obtain an initial feature vector set, and performing cluster analysis and topic modeling on the initial feature vector set to obtain a data topic distribution matrix; Based on the data topic distribution matrix, semantic similarity calculation and association network construction are performed to obtain a data association graph, and centrality analysis and importance evaluation are performed on the data association graph to obtain a data node weight list; Based on the data node weight list and the predefined sensitive word library, the sensitive information of the enterprise data is identified and quantitatively scored to obtain an initial sensitivity score, and according to the initial sensitivity score, time decay modeling and dynamic weight calculation are performed on the enterprise data to obtain a sensitivity index; The enterprise data is classified at multiple levels and labels are generated according to the sensitivity index to obtain an initial classification labeling result, and the initial classification labeling result is used to perform classification integration and metadata encapsulation to obtain classification labeling data with a sensitivity index.
3. The enterprise sensitive data security access management method according to claim 1 is characterized in that: The step of performing association analysis based on the classification tag data and user attribute information to obtain an adaptive access control matrix includes: Performing feature vectorization processing on user attribute information to obtain a user feature matrix, and performing user role identification based on the user feature matrix to obtain an initial role division result; Based on the initial role division result and the classified label data, association rule mining is performed on the user-data access relationship to obtain frequent item sets and association rule sets, and bipartite graph construction and weight assignment are performed on user roles and data categories according to the frequent item sets and the association rule sets to obtain a role-data association network; Calculating user role centrality and ranking importance based on the role-data association network to obtain a role weight vector, and quantifying the authority level of the role-data association according to the role weight vector and the sensitivity index to obtain an initial authority matrix; Based on the initial permission matrix, correlation analysis is performed on permission allocation to obtain a permission rule set, and conditional probability calculation and Bayesian network construction are performed on the initial permission matrix according to the permission rule set to obtain a dynamic permission probability distribution; Based on the dynamic permission probability distribution, permission allocation time series prediction and anomaly detection are performed to obtain a permission adjustment suggestion list, and based on the permission adjustment suggestion list, permission allocation multi-objective optimization and matrix reconstruction are performed to obtain an adaptive access control matrix.
4. The enterprise sensitive data security access management method according to claim 1 is characterized in that: The method of performing all-round recording and pattern analysis on the user's access to the classified tagged data according to the adaptive access control matrix to obtain a data access audit log includes: Capturing user access behaviors in real time and serializing events to obtain an original access event stream, and performing permission verification and compliance checks on the user access behaviors based on the original access event stream and the adaptive access control matrix to obtain access compliance marking results; Based on the access compliance marking results, the access events are subjected to spatiotemporal feature extraction and behavior vectorization to obtain a multi-dimensional behavior feature matrix, and the user access patterns are analyzed and typical behaviors are identified based on the multi-dimensional behavior feature matrix to obtain a behavior pattern prototype set; Based on the behavior pattern prototype set and historical access data, behavior time series modeling and trend prediction are performed to obtain a behavior evolution trajectory diagram, and behavior deviation calculation and anomaly scoring are performed based on the behavior evolution trajectory diagram and predefined anomaly indicators to obtain a behavior anomaly measurement scale; Based on the behavior anomaly measurement table and the sensitivity index, behavior risk level assessment and priority sorting are performed to obtain a risk event queue, and audit task scheduling optimization and parallel processing are performed through the risk event queue to obtain a distributed audit task graph; Based on the distributed audit task graph, audit data is incrementally aggregated and compressed to obtain structured audit log entries, and audit information association analysis and causal reasoning are performed based on the structured audit log entries to obtain a data access audit log.
5. The enterprise sensitive data security access management method according to claim 1 is characterized in that: The integrated analysis based on the data access audit log, the decryption authority policy, the access record of the encrypted data and the external threat information is performed to obtain a dynamic defense strategy and a real-time response plan, and the adaptive access control matrix is updated according to the dynamic defense strategy to obtain a target access control matrix, including: Correlation and pattern mining are performed on the data access audit log and the access record of the encrypted data to obtain a user behavior sequence diagram, and user operation compliance assessment and anomaly detection are performed based on the user behavior sequence diagram and the decryption authority policy to obtain a behavior anomaly measurement table; Perform security event correlation analysis based on the behavior anomaly measurement table and external threat information to obtain a threat event correlation network, and perform potential attack path reasoning and probability estimation based on the threat event correlation network to obtain an attack path probability tree; Based on the attack path probability tree, defense resources are optimally allocated and deployed to plan, a preliminary defense strategy is obtained, and according to the preliminary defense strategy, cost-benefit evaluation and priority ranking of defense measures are performed to obtain a hierarchical defense strategy set; Dynamically generate defense rules and detect conflicts based on the hierarchical defense strategy set to obtain a target defense rule base, and automatically orchestrate emergency response processes and build a decision tree based on the target defense rule base to obtain a real-time response plan; Based on the real-time response scheme, dynamic adjustment of access rights and update of constraint conditions are performed to obtain a temporary access control strategy, and according to the temporary access control strategy, global optimization and balanced adjustment are performed on the adaptive access control matrix to obtain a target access control matrix.
6. An enterprise sensitive data security access management system, characterized in that: The system is used to execute the enterprise sensitive data security access management method according to any one of claims 1 to 5, and comprises: The evaluation module is used to conduct multi-dimensional analysis and dynamic evaluation of enterprise data to obtain classified labeled data with sensitivity index; An association module, used for performing association analysis based on the classification mark data and user attribute information to obtain an adaptive access control matrix; An analysis module, used to comprehensively record and analyze patterns of user access behaviors of the classified labeled data according to the adaptive access control matrix to obtain a data access audit log; A configuration module, for performing security domain division and collaboration strategy configuration on the classified and marked data based on the data access audit log, to obtain a virtual collaboration space and data sharing rules; A processing module, configured to perform differential encryption processing and key generation on the classified labeled data according to the sensitivity index, the virtual collaborative space and the data sharing rule, to obtain encrypted data and a decryption authority policy; An update module is used to perform integrated analysis based on the data access audit log, the decryption permission policy, the access record of the encrypted data and the external threat information to obtain a dynamic defense strategy and a real-time response plan, and to update the adaptive access control matrix according to the dynamic defense strategy to obtain a target access control matrix.
7. An enterprise sensitive data security access management device, characterized in that: The enterprise sensitive data security access management device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instruction in the memory to enable the enterprise sensitive data security access management device to execute the enterprise sensitive data security access management method as described in any one of claims 1-5.
8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the enterprise sensitive data security access management method as described in any one of claims 1-5 is implemented.
