Collaborative management permission rule optimization system based on community discovery

Abstract

A collaborative management permission rule optimization system based on community discovery includes a feature extraction module, an organization role screening module, a data attribute screening and permission allocation control module, the feature extraction module generates standardized user feature vectors and data vectors through structured coding and semantic modeling according to user data, collaboration objects and data attribute information; the organization role screening module constructs a user collaboration graph based on a community division algorithm and infers the optimal role mapping of the user using a community portrait model; the data attribute screening module identifies the semantic category and upstream and downstream dependency chain of the target data in combination with a data classification tree and a structure dependency relationship extraction algorithm; the permission allocation control module infers in combination with a permission knowledge base according to the role, data classification and dependency relationship, and finally generates a permission allocation table, realizing dynamic permission configuration and strategy matching for collaborative scenarios. The present application realizes precise permission allocation and real-time strategy adjustment in the ship collaborative scenario through automatic role mapping, data semantic association analysis and permission inheritance reasoning, thereby improving data security and collaboration efficiency and providing core support for the digital transformation of the ship industry.

Description

Technical Field

[0001] This invention relates to a technology in the field of shipbuilding, specifically a collaborative management permission rule optimization system based on community discovery. Background Technology

[0002] Static permission models are difficult to adapt to multi-level collaborative scenarios in the shipbuilding industry. They cannot dynamically perceive organizational structure evolution and data dependencies, resulting in broken permission inheritance paths, hindered cross-project and cross-organization sharing, and difficulty in meeting the fine-grained access requirements for 3D model review. Summary of the Invention

[0003] To address the aforementioned shortcomings of existing technologies, this invention proposes a collaborative management permission rule optimization system based on community discovery. Through automated role mapping, data semantic association analysis, and permission inheritance reasoning, it achieves accurate permission allocation and real-time policy adjustment in ship collaboration scenarios, thereby improving data security and collaboration efficiency and providing core support for the digital transformation of the shipbuilding industry.

[0004] This invention is achieved through the following method:

[0005] This invention relates to a collaborative management permission rule calculation system based on community discovery, comprising: a feature extraction module, an organization role filtering module, a data attribute filtering module, and a permission allocation control module. Specifically: the feature extraction module generates standardized user feature vectors and data vectors based on user data, collaborating objects, and data attribute information through structured encoding and semantic modeling; the organization role filtering module constructs a user collaboration graph based on a community segmentation algorithm and uses a community profiling model to infer the optimal user role mapping; the data attribute filtering module combines data classification trees and structural dependency extraction algorithms to identify the semantic categories and upstream / downstream dependency chains of target data; and the permission allocation control module, based on roles, data classifications, and dependencies, and combined with a permission knowledge base, performs reasoning to ultimately generate a permission allocation table, achieving dynamic permission configuration and strategy matching for collaborative scenarios.

[0006] The described community profiling model employs a layered design comprising a community layer, an organizational role layer, and a data layer. The community layer contains multiple users who form a collaborative network through project collaboration and task coordination. The organizational role layer, based on a collaboration graph structure, uses a dual-classification mechanism of role-based communities and organizational communities to identify highly cohesive user groups from the perspectives of functional attributes and collaborative relationships. It generates community profiles through feature aggregation and dynamically maps the current user to the optimal collaborative role using similarity calculation and path behavior reasoning. The data layer comprises accessible data resources, including folders, files, BOM structures, and version information, supporting fine-grained access control. By integrating the classification results of role-based communities and organizational communities, this model achieves a multi-dimensional and accurate mapping of users, roles, and data, providing dynamic permission allocation strategies for complex collaborative scenarios.

[0007] The feature extraction refers to extracting multi-dimensional features from users, collaborative scenarios, and data objects, and constructing highly discriminative feature sets through structured encoding, semantic modeling, and vectorization techniques to provide standardized input for subsequent permission calculations. This process employs a combination of metadata parsing, graph embedding, and context encoding.

[0008] The aforementioned organizational role selection refers to: realizing dynamic role mapping based on user feature vectors, mining user collaboration patterns through community discovery technology, and deriving the optimal role allocation scheme by combining multi-dimensional similarity calculations.

[0009] The community profiling model jointly models the consistency of collaborative behavior, the semantic consistency of roles, and the semantic consistency of organizations. Each community node simultaneously carries three types of homogeneous information: role-role similarity, organization-organization similarity, and their corresponding collaboration patterns, thereby expanding the traditional single collaboration graph into a role-organization dual-channel heterogeneous graph. Specifically, it is obtained through the following method:

[0010] a. Graph structure initialization: Constructing a heterogeneous graph With users as nodes The node attributes include the character vector. Organizational vectors And a project list; constructing three types of edges: role similarity edges. weight =cos( , ), organizing similar edges weight =cos( , ), Collaborative Edge weight ,in: , , These are weighting factors for the number of joint projects, communication frequency, and most recent collaboration time, respectively.

[0011] b. Community segmentation: In The Louvain hierarchical detection is performed in parallel twice: the first time with To obtain the character community collection for weighting The second time The weighted set of the organizational community .

[0012] c. Community Profile Generation: For any character's community With the corresponding organization community The role center is obtained by aggregating the member role vector and the organization vector respectively. With the organization center And extract high-frequency role tags. Organizational tags and typical collaborative paths (g); ultimately forming a dual profile. This is used for subsequent double similarity calculations and permission condition merging.

[0013] The community similarity mentioned is calculated within the role-organization dual-channel heterogeneous graph framework, targeting role communities separately. With the organization community Obtained independently. Specifically, it is obtained through the following method:

[0014] a. Vector space similarity: Calculating user feature vectors With each community's central vector Cosine similarity: This reflects the degree of matching between users and static attributes of roles, organizations, and projects.

[0015] b. Behavioral path distance: based on user action sequence Typical Community Path The behavioral pattern differences are calculated using Dynamic Time Warping (DTW): The negative sign indicates that the smaller the distance, the higher the similarity, which is used to capture dynamic collaborative features.

[0016] c. Collaboration frequency overlap rate: Statistics of users Overlap with actual collaboration records of community members: ,in: For users With community The similarity score in the history of collaboration indicates that the higher the value, the closer the social relationship. For users The historical collection of collaborating users, that is, all users who have collaborated with the user A collection of other user IDs who have participated in the project or task together; For the community A collection of all historical collaborating users within the group.

[0017] d. Comprehensive Similarity Calculation: By weighted fusion of multi-dimensional similarity indicators, user similarity is defined. With community The overall similarity is: ,in: For users With community The overall similarity score is calculated. Finally, based on the overall similarity score, the system outputs a comparison with the user's score. A list of the K most similar communities: ,in: The first community in the candidate community set C A community, For the first The normalized comprehensive similarity score of each community. This comprehensive scoring mechanism realizes a unified expression of multi-dimensional similarity, enabling individual users to quickly locate the community groups with the most similar behavioral patterns, structural attributes, and collaborative tendencies in complex organizational collaborative networks.

[0018] The aforementioned data attribute filtering refers to the automatic parsing and aggregation of the attributes, categories, and structural relationships of the accessed data objects, and the generation of data context labels that can be used for permission reasoning through multi-level semantic classification, dependency extraction, and dynamic constraint verification.

[0019] The dependency extraction mentioned above includes:

[0020] a. Structure Graph Construction. Based on the BOM structure and hierarchical directory, construct a directed graph. Where: the node set V is all data entities with permission management significance, and the edge set E is the structural dependencies, reference relationships, or hierarchical subordinate relationships between data objects. The directed graph G serves as the basic framework for dependency extraction.

[0021] b. Target node identification. Receive the data object identifier involved in the user's current operation request. Locate the corresponding node in the graph structure G. This node, serving as the starting point for dependency path traversal, is the key entry point for subsequently extracting its upstream and downstream related data. This step supports precise location of data objects using unique identifiers (such as file IDs, document numbers, and object paths).

[0022] c. Traversing upstream and downstream paths. A depth-first search (DFS) algorithm is used to traverse the path from the target node. Starting from this point, recursively traverse all its downstream and upstream nodes, recording all reachable paths: .

[0023] The dynamic constraint check mentioned above refers to: based on the dependency legality condition: ,in: , For dependency paths The declared start and end objects, The first one in the user operation log sequence The object and conditions affected by this actual operation This is used to ensure that the order of operations is consistent with the direction of dependencies. After the above judgment, the set of valid dependencies for the context is filtered out: .

[0024] The aforementioned permission allocation and control refers to: based on the user's optimal role mapping result, the classification tags of the target data and their dependencies, and combined with a preset permission rule knowledge base, comprehensively inferring the user's operable permissions on a specific data object.

[0025] The classification labels and their dependencies of the target data are obtained in the following way:

[0026] a. Rule matching and filtering: Based on the user's optimal role mapping results Data classification tags and the set of legal dependency paths In the permission knowledge base Perform multi-dimensional condition filtering to select all permission rule items that meet the conditions. Among them: requirements rules Character Collection Optimal set of user roles A subset of, and rules Data Classification Category labels of target data objects Totally consistent.

[0027] b. Generate a permission set: For rule items that match successfully... Extract the permission set field from each rule item. Perform a set union operation on the permission sets of all matching items to form the initial inherited permission set, specifically: This set represents the operational permissions that a user can directly infer based on roles and data tags in the current context, and it forms the basis for subsequent dependency extension logic.

[0028] c. Dependency chain permission expansion: In the initial permission set Based on this, a depth-first search (DFS) traversal is used. Each dependency path in Recursively check each node on the path Permission inheritance requirements: If node If the corresponding permission rule is marked as transitive, the relevant permissions on its parent node are passed to the current node through the inheritance mechanism, and the transmission path and inheritance source are recorded. If an edge in the dependency path is defined as a strong constraint, this step will simultaneously verify whether the permission set of the predecessor node on the path contains the prerequisite permission set required by the edge.

[0029] The standard permission allocation entries include user ID, role name, data object ID, data category, type of permission granted (such as read, write, copy, delete, approval), scope of permission, and source of permission.

[0030] Technical effect

[0031] This invention introduces a community discovery algorithm into permission management, transforming the originally static role configuration into a dynamic mapping that evolves in real time with organizational structure and collaborative behavior. This eliminates the need for manual maintenance by administrators, ensuring accurate matching of roles and permissions in a multi-project parallel shipbuilding environment. Simultaneously, it utilizes graph-driven data semantic parsing to uniformly encode files, BOM nodes, and their upstream and downstream dependencies as context labels. This allows permission rules to reason based on fine-grained data structures and semantic relationships, avoiding redundant authorizations and permission omissions. Furthermore, an incremental graph update mechanism ensures that when personnel, projects, or data relationships change, permission adjustments only require local recalculation, significantly improving the scalability of collaborative management permission rule calculation methods. The final standardized permission allocation table can be seamlessly integrated into existing systems, enabling fine-grained secure collaboration across organizations and overcoming the application limitations of traditional RBAC models in dynamic collaborative scenarios. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the structure of the present invention;

[0033] Figure 2 A schematic diagram of a community profiling model;

[0034] Figure 3 This is a schematic diagram of an embodiment. Detailed Implementation

[0035] like Figure 1 As shown in the figure, this embodiment relates to a collaborative management permission rule calculation system based on community discovery, which includes: a feature extraction module, an organization role filtering module, a data attribute filtering and permission allocation control module.

[0036] The feature extraction module includes: a user data parsing unit, a collaborative scenario modeling unit, a data attribute parsing unit, and a feature standardization processing unit. Specifically: the user data parsing unit collects user information, including role descriptions, organization identifiers, and project tags, combining these multi-dimensional fields to generate user vectors; the collaborative scenario modeling unit extracts operation type, collaborative object type, and access frequency temporal features based on user operation logs. Context modeling is performed using a lightweight Transformer encoder to generate dynamic collaborative semantic vectors; the data attribute parsing unit analyzes the type identifier, version number, and organization attribute in the metadata of the accessed data object. A hierarchical embedding strategy is used: structured attributes are encoded using regularization, while unstructured attributes are extracted using a BERT fine-tuning model to extract semantic features. The final generated multi-dimensional vector of the data object is: The feature standardization processing unit processes user vectors. and data vector Normalization was performed separately to eliminate dimensional differences. After mapping the eigenvalues ​​of each dimension to the [0,1] interval using Min-Max Scaling, principal component analysis (PCA) was used to compress the high-dimensional vector to a uniform dimension. To reduce computational complexity, one-hot encoding is used to enhance the separability of discrete features, and the final output is a standardized feature vector: as well as This ensures that the similarity calculation and cluster analysis of subsequent modules have numerical stability.

[0037] The aforementioned combination of multidimensional fields generates user vectors by uniformly encoding role descriptions, organization identifiers, and project tags to construct a standard input vector. An entity recognition method based on rule templates or keyword dictionaries is then used to identify role and organization entities, forming triples. A multi-level mapping mechanism is used to establish the association between roles and permission models. Finally, multi-dimensional fields are combined to generate user vectors. This is used to support subsequent permission matching and modeling.

[0038] The organizational role selection module includes: a community profile model construction unit, a community similarity calculation unit, a role-organization mapping reasoning unit, and an optimal role selection unit. Specifically: the community profile model construction unit constructs a community profile model G based on historical collaborative behavior data of the organization; the community similarity calculation unit calculates the similarity between the current user's feature vector and the community nodes in the community profile model G, using vector space similarity, behavioral path distance, and collaborative frequency overlap rate as metrics to calculate community similarity. The role-organization mapping inference unit extracts role tags from Top-K communities and generates a candidate role probability distribution by weighting the distribution based on community similarity. Output the set of candidate characters sorted by probability. The optimal role selection unit, combining organizational constraints and project requirements, outputs all roles that meet the criteria from the candidate set R. Final output. Multiple role mapping relationships enable dynamic role allocation.

[0039] The data attribute filtering module includes: a data classification and clustering unit, a classification tree matching unit, a data dependency extraction unit, and a dynamic dependency verification unit, wherein: the data classification and clustering unit receives the data attribute vector input from the feature extraction module. Extract key attribute set Generate attribute vector sets through semantic embedding methods The semantic similarity between any two attribute pairs is calculated, and the DBSCAN clustering algorithm is used to classify the attributes into semantically consistent categories, resulting in a preliminary set of categorized nodes. The classification tree matching unit uses a built-in system of nodes containing multiple pre-trained semantic vectors. Data classification tree Calculate the candidate class node vector The cosine similarity with the vector of each node in the classification tree is used. The class node with the highest similarity is selected as the standard classification label for the data object, which is used as the basis for subsequent permission matching and rule association. This process ensures that unstructured or heterogeneous data can be normalized in permission management. The data dependency extraction unit constructs a directed graph to represent the dependencies between data objects and traverses the graph using a depth-first search (DFS) algorithm to identify the upstream and downstream dependency paths of the target data object. The dynamic dependency verification unit combines user operation sequences... With dependency path Verify whether the dependency path is valid in the actual operation context.

[0040] The permission allocation and control module includes: a permission knowledge base construction unit, a dependent permission inheritance unit, an explicit authorization overriding unit, and a permission allocation table construction unit, wherein: the permission knowledge base construction unit extracts structured permission rule items from historical authorization logs and system policy templates. The rules are summarized and organized into a permission knowledge base. This is used for subsequent permission matching and inference; it depends on the permission inheritance unit based on the user's optimal role mapping result. Data classification tags and the set of legal dependency paths Perform rule matching and filtering to generate a permission set, and then extend the dependency chain permissions based on this set; the explicit authorization overriding unit queries the explicit authorization record table G, and if there is a user's manual authorization record for the target object, it checks if such a record exists. Then grant explicit permissions. Override the inheritance result to obtain the final permission set. The permission allocation table construction unit is based on the user's optimal role. The associated project and organizational information, the unique identifier of the target data object, and its classification results. Each authorization item is output as a standard permission allocation entry, forming the final permission allocation table of the system.

[0041] like Figure 3 As shown, the collaborative management permission rule management device based on the above system includes: a web application layer, a community discovery-based collaborative management permission rule calculation system as the business processing layer, and a data layer. The web application layer implements the user interface and system interaction functions, supporting permission rule configuration, operation log monitoring, and the display and export of permission results. The community discovery-based collaborative management permission rule calculation system integrates multiple functional modules such as feature extraction, organizational role filtering, data attribute filtering, and permission allocation control, completing the entire process calculation from user data input, feature modeling, permission inference to permission allocation table output. The data layer carries all permission-related multimodal data, including structured user information, unstructured metadata, graph-structured collaborative networks, and dependency paths, supporting unified management and efficient retrieval of different types of data. The entire system operates collaboratively across three layers, achieving automated, dynamic, and refined permission allocation in collaborative scenarios, adapting to complex and ever-changing project organizational relationships and task structures.

[0042] The web application layer is implemented based on the Vue.js framework and uses the ElementPlus component library to handle permission rule configuration, permission status display, and interactive form design. The system allows users to upload JSON files containing user information, collaborative scenario information, and data object information via the front-end interface. After parsing, the front-end initiates a permission calculation request to the back-end FastAPI service via Axios. During the request process, RESTful API parameters are transmitted and structure is validated, and WebSocket is used to monitor the calculation status and result feedback. After permission calculation is complete, the result is returned to the front-end in standardized JSON format, and users can choose to export it as a CSV file for subsequent analysis. The system front-end also integrates a permission rule template management function, supporting a visual display of the mapping relationship between users, data objects, and permissions, and allowing users to add, delete, and modify rules.

[0043] The business processing layer is the core functional module of the system, completing all computational logic from raw input to final permission output. The system first parses user information and data object information, calling the feature extraction module to generate user feature vectors and data feature vectors. Then, it constructs a collaboration graph structure, combining collaboration frequency and communication behavior from historical operation logs to generate a heterogeneous graph structure in Neo4j, and uses the Louvain algorithm to segment communities along the role and organization dimensions, forming role-based and organization-based community profiles. The system infers the optimal role set for the current user based on vector similarity and path overlap indices. In the data processing stage, the system matches classification tree nodes based on data vectors, extracts the upstream and downstream paths of the data object in the dependency graph, and determines the legality of dependency paths based on the user's actual operation sequence. The permission control module matches corresponding rules in the permission knowledge base based on the optimal role, data classification labels, and legal dependency paths, generating a permission set and outputting structured results.

[0044] The data layer employs a collaborative working mode using three database types: PostgreSQL, MongoDB, and Neo4j. These manage structured information, document-based permission data, and graph-based dependencies, respectively. PostgreSQL stores user information, role attributes, organizational relationships, and metadata of data objects, supporting transactional queries and conditional filtering. MongoDB stores permission rule items, user authorization records, and intermediate results in document format, suitable for various field structures and supporting conditional retrieval by project or user dimension. The Neo4j graph database records user collaboration graphs and data dependency graphs. The system uses a graph query language to call built-in DFS and Louvain algorithms to complete dependency path extraction and community discovery operations. The system's data layer structure supports incremental updates and concurrent access. Each permission request from the front end triggers a backend query of the three data sources, providing data support for the entire process from feature extraction to rule matching.

[0045] Table 1 Comparison of Technical Features

[0046]

[0047] This invention, through the synergistic integration of multiple core algorithms and key technical features, significantly outperforms existing permission management methods in terms of completeness, accuracy, and flexibility. Regarding completeness, Transformer temporal modeling in feature extraction fully captures dynamic user behavior features; Louvain community segmentation in organizational role selection comprehensively covers organizational collaboration relationships; and DBSCAN clustering and Neo4j graph analysis in data attribute selection construct semantic associations for data, forming a three-dimensional basis for permission decision-making. In terms of accuracy, organizational role selection employs a triple-weighted calculation based on feature similarity, behavioral path matching, and collaboration frequency overlap, combined with dependency path verification and explicit authorization coverage mechanisms for permission allocation. This ensures accurate output of user permission sets even in multi-role scenarios. Regarding flexibility, organizational role selection utilizes the incremental calculation function of the Louvain algorithm to achieve local graph reconstruction. When a change in user organizational relationships is detected, the algorithm automatically triggers a partial update of the community division instead of a full recalculation. Data attribute selection leverages the real-time indexing mechanism of the Neo4j graph database to dynamically adjust the data dependency network, enabling the collaborative management permission rule calculation method to automatically adapt to typical multi-role change scenarios in the shipbuilding industry, addressing the failure of traditional manual maintenance methods in large-scale, complex organizations. This invention not only effectively solves the problem of fine-grained permission control in cross-organizational collaboration scenarios but also provides an innovative solution for designing a secure, reliable, and scalable collaborative management permission rule optimization system, possessing broad application value and significant technical advantages.

[0048] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A collaborative management permission rule calculation system based on community discovery, characterized in that, include: The system comprises a feature extraction module, an organizational role selection module, a data attribute selection module, and a permission allocation control module. Specifically: the feature extraction module generates standardized user feature vectors and data vectors based on user data, collaborating objects, and data attribute information through structured coding and semantic modeling; the organizational role selection module constructs a user collaboration graph based on a community segmentation algorithm and uses a community profiling model to infer the optimal user role mapping; the data attribute selection module combines a classification tree and a data dependency extraction algorithm to identify the semantic category and upstream / downstream dependency chains of the target data; the classification tree is a predefined tree structure containing multiple nodes with pre-trained semantic vectors; identifying the semantic category of the target data refers to calculating the cosine similarity between the attribute vector of the target data and the vector of each node in the classification tree, and selecting the one with the highest similarity. The class nodes are used as standard classification labels; the identification of upstream and downstream dependency chains refers to: constructing a directed graph between data objects and traversing the graph using the depth-first search (DFS) algorithm to determine the upstream and downstream dependency paths of the data; the permission allocation control module, based on roles, data classifications, and dependencies, and combined with the permission knowledge base, performs reasoning to finally generate a permission allocation table, realizing dynamic permission configuration and policy matching for collaborative scenarios; the permission knowledge base contains structured permission rule items extracted from historical authorization logs and system policy templates; the reasoning refers to: filtering rule items in the permission knowledge base that meet the conditions, namely, the set of roles in the rule is a subset of the user's optimal set of roles and the data classification of the rule is consistent with the target data, and performing a set union operation on the permission sets of all matching items; The community profiling model adopts a layered design of community layer, organizational role layer, and data layer. The community layer contains multiple system users who form a collaborative network through project collaboration and task coordination. The organizational role layer is based on a dual-channel heterogeneous graph structure containing role similarity edges, organizational similarity edges, and collaborative edges. It uses a dual partitioning mechanism of role communities and organizational communities based on user role attributes and organizational attributes, identifying highly cohesive user groups from the perspectives of functional attributes and collaborative relationships. The user's role attributes and organizational attributes correspond to role vectors and organizational vectors in the heterogeneous graph nodes, respectively. The highly cohesive user groups refer to the set of role communities and the set of organizational communities partitioned using the Louvain hierarchical detection algorithm, based on role similarity edge weights and organizational similarity edge weights, respectively. A community profile is generated through feature aggregation, and the current user is dynamically mapped to the optimal collaborative role by combining similarity calculation and path behavior reasoning. The combination of similarity calculation and path behavior reasoning refers to calculating the cosine similarity between the user feature vector and the community center vector, and using Dynamic Time Warping (DTW) to calculate the user's cosine similarity. Actual operation log sequence Typical Community Path Differences in behavioral patterns ;in Represents a specific community The model defines the standard behaviors or collaboration patterns typically exhibited by members; the optimal collaborative role refers to the role label in the community profile that best matches the current user's characteristics and behavioral patterns; the data layer comprises the data resources accessible in the system, including folders, files, BOM structure, and version information, supporting fine-grained operation permission control. This model integrates the division results of role communities and organizational communities to achieve multi-dimensional and accurate mapping of users, roles, and data, providing dynamic permission allocation strategies for complex collaboration scenarios. The data attribute filtering module includes: a data classification and clustering unit, a classification tree matching unit, a data dependency extraction unit, and a dynamic dependency verification unit, wherein: the data classification and clustering unit receives the data attribute vector input from the feature extraction module. Extract key attribute set Generate attribute vector sets through semantic embedding methods The semantic similarity between any two attribute pairs is calculated, and the DBSCAN clustering algorithm is used to classify the attributes into semantically consistent categories, resulting in a preliminary set of classification nodes. The classification tree matching unit uses a built-in node containing multiple pre-trained semantic vector representations. Classification tree Calculate the candidate class node vector The class node with the highest cosine similarity to the vector of each node in the classification tree is selected as the standard classification label for the candidate class node vector. This label is used as the basis for subsequent permission matching and rule association to ensure that unstructured or heterogeneous data can be normalized in permission management. The data dependency extraction unit constructs a directed graph to represent the dependencies between data objects and traverses the graph using a depth-first search (DFS) algorithm to identify the upstream and downstream dependency paths of the target data object. The dynamic dependency verification unit obtains the user's actual operation log sequence. With dependency path Verifying the validity of dependency paths within the actual operational context involves comparing the order of operations on the start and end objects in the dependency path with the timestamps of the actual operations in the user logs. A dependency is considered valid only if the actual order of operations matches the direction of the dependency. The permission allocation and control module includes: a permission knowledge base construction unit, a dependent permission inheritance unit, an explicit authorization overriding unit, and a permission allocation table construction unit, wherein: the permission knowledge base construction unit extracts structured permission rule items from historical authorization logs and system policy templates. ,in Represents a set of characters. Indicates data classification, Represents a set of conditions. This represents the set of granted permissions, and the rules are summarized and organized into a permission knowledge base. This is used for subsequent permission matching and inference; it depends on the permission inheritance unit based on the user's optimal role mapping result. Data classification tags and the set of legal dependency paths The system performs rule matching and filtering to generate a permission set, and then extends the dependency chain permissions based on this set. The explicit authorization overriding unit queries the explicit authorization record table; if a user's manual authorization record for the target object exists, then the explicit authorization permission is applied. Override the inheritance result to obtain the final permission set. The permission allocation table construction unit is based on the user's optimal role. The optimal role is associated with the project and organization information, the unique identifier of the target data object, and the data classification label of the target data object. Each authorization item is output as a standard permission allocation entry, forming the final permission allocation table of the system.

2. The collaborative management permission rule calculation system based on community discovery according to claim 1, characterized in that, The feature extraction module includes: a user data parsing unit, a collaborative scenario modeling unit, a data attribute parsing unit, and a feature standardization processing unit. Specifically: the user data parsing unit collects user role descriptions, organization identifiers, and project tags from user information, combining these multi-dimensional fields to generate user vectors; the collaborative scenario modeling unit extracts operation type, collaborative object type, and access frequency temporal features from user operation logs, and performs context modeling using a lightweight Transformer encoder to generate dynamic collaborative semantic vectors; the data attribute parsing unit analyzes the type identifier, version number, and organization attribute in the metadata of the accessed data object, and uses a hierarchical embedding strategy: structured attributes are encoded using regularization, and unstructured attributes are extracted using a BERT fine-tuning model to extract semantic features, encoding the type identifier, version number, and organization attribute into type feature vectors. Version feature vector and organizational attribute feature vector Ultimately, this is achieved through vector concatenation operations. Generate multidimensional vectors of data objects The feature standardization processing unit processes user vectors. and data vector Normalization was performed separately to eliminate dimensional differences. After mapping the eigenvalues ​​of each dimension to the [0,1] interval using Min-Max Scaling, principal component analysis (PCA) was used to compress the high-dimensional vector to a uniform dimension. To reduce computational complexity, one-hot encoding is used to enhance the separability of discrete features, and the final output is a standardized feature vector: as well as This ensures that the similarity calculation and cluster analysis of subsequent modules have numerical stability; The hierarchical embedding strategy refers to the following: structured attributes are encoded using regularization, while unstructured attributes are extracted using a BERT fine-tuning model to extract semantic features.

3. The collaborative management permission rule calculation system based on community discovery according to claim 2, characterized in that, The aforementioned combination of multidimensional fields generates user vectors by uniformly encoding role descriptions, organization identifiers, and project tags to construct a standard input vector. Then, an entity recognition method based on rule templates or keyword dictionaries is used to identify role and organization entities, forming triples. ,in A unique identifier representing a user entity. This represents the user's ownership of roles. A unique identifier representing a role; a multi-level mapping mechanism establishes the association between roles and the permission model, and finally, multi-dimensional fields are combined to generate a user vector. This is used to support subsequent permission matching and modeling; The multi-level mapping mechanism includes: a first-level mapping, which maps user entities to initial role identifiers based on the triples; a second-level mapping, which dynamically maps the initial role identifiers to the optimal collaborative role in the community profile by combining the community profile model and similarity calculation; and a third-level mapping, which maps the optimal collaborative role to the set of operation permissions in the permission model based on the permission knowledge base.

4. The collaborative management permission rule calculation system based on community discovery according to claim 1, characterized in that, The organizational role selection module includes: a community profile model construction unit, a community similarity calculation unit, a role-organization mapping reasoning unit, and an optimal role selection unit. Specifically: the community profile model construction unit constructs a community profile model G based on historical collaborative behavior data of the organization; the community similarity calculation unit calculates the similarity between the current user's feature vector and the community nodes in the community profile model G, using vector space similarity, behavioral path distance, and collaborative frequency overlap rate as metrics to calculate community similarity. The role-organization mapping inference unit extracts role tags from Top-K communities and generates a candidate role probability distribution by weighting the distribution based on community similarity. ,in Indicates the first Given 10 candidate characters, output the set of candidate characters sorted by probability. ,in to Indicates candidate roles, to This represents the corresponding probability value; the optimal role selection unit, combining organizational constraints and project requirements, outputs the optimal role mapping results (roles) for all users meeting the conditions from the candidate set R, and finally outputs... Multiple role mapping relationships enable dynamic role allocation.

5. A collaborative management permission rule management device for a system according to any one of claims 1-4, characterized in that, include: The system consists of a web application layer, a community-based collaborative management permission rule calculation system, and a data layer. The web application layer is used for user-system interaction. The business processing layer, implemented through the collaborative management permission rule optimization system, comprehensively processes user characteristics, collaborative relationships, and data dependencies. The collaborative management permission rule calculation system, as the business processing layer, completes dynamic role mapping, permission boundary deduction, and optimized output of the permission allocation table, generating data calculation and dynamic permission rules for the entire process from user input to permission output. The data layer optimizes storage and enables efficient access by classifying multimodal data characteristics for permission management scenarios.

Images