Standard electronic archive management method and system based on artificial intelligence

Abstract

The invention discloses a standard electronic archive management method and system based on artificial intelligence, and the method comprises the steps: carrying out the inspection of four properties of electronic archives, dividing the electronic archives into high-quality, qualified and problematic archives, learning the file features of the archives, distributing data streams, analyzing the access behaviors of a user to the data streams of various archives, predicting the next archive, and dynamically adjusting the authority according to the predicted next archive. The method comprises the following steps: constructing a knowledge graph according to electronic archive semantic feature association and user permission, predicting a risk propagation range based on graph node information, temporarily isolating and backing up an affected archive, chaining and storing the electronic archive based on a block chain technology, and recording each access and modification operation on the electronic archive to form a complete operation auditing chain. According to the method, archive resources can be efficiently integrated, the security and credibility of archive management are enhanced, a comprehensive, reliable and intelligent solution is provided for electronic archive management, and meanwhile the method has good interpretability.

Description

Technical Field

[0001] The present invention relates to the field of file management, and in particular to a standard electronic archive management method and system based on artificial intelligence. Background Art

[0002] With the rapid development of information technology, electronic archives have become a core component of information management in various industries. In today's digital age, enterprises, government agencies and various organizations generate massive amounts of electronic documents every day, such as contract documents, financial statements, project documents, administrative approval records, etc. These electronic archives contain rich information resources and are of vital importance to the organization's operational management, decision-making, compliance auditing and historical data preservation.

[0003] Traditional electronic archive management methods mainly rely on manual operations and simple database management systems. In the archive collection process, manual collection is inefficient and error-prone, and it is difficult to cope with large-scale, multi-source data aggregation. In terms of archive classification and storage, traditional methods are mostly based on pre-set fixed directory structures and keyword retrieval. This method lacks flexibility and intelligence, and cannot accurately understand the semantics of archive content.

[0004] An innovative standard electronic archive management method and system based on artificial intelligence, which realizes the intelligent, efficient, precise and all-round "four-quality" guarantee of electronic archive management, improves the organization's information management level and core competitiveness, and adapts to the requirements of the rapid development of the digital age. Summary of the invention

[0005] The purpose of the present invention is to provide a standard electronic archive management method and system based on artificial intelligence.

[0006] To achieve the above object, the present invention is implemented according to the following technical solutions: The first aspect of the present invention provides a standard electronic archive management method based on artificial intelligence, comprising: Conducting four-property inspections on electronic archives, classifying the electronic archives into high-quality archives, qualified archives and problematic archives according to the inspection results, learning the file characteristics of various electronic archives, and allocating data streams for various electronic archives; the four-property inspections include authenticity, integrity, availability and security inspections; Analyze the user's access behavior to various archive data streams, predict the next archive the user will access based on the analysis results, and dynamically adjust the user's access rights using reinforcement learning based on the prediction results; Building a knowledge graph based on the association of semantic features of electronic archives and the access rights of the users; According to the location, connectivity and information propagation path of the nodes in the knowledge graph, the scope of risk propagation is predicted, and the affected electronic files within the scope are temporarily isolated and backed up; Based on blockchain technology, electronic files are stored on the chain, and every access and modification operation of the electronic files is recorded on the blockchain to obtain a complete operation audit chain.

[0007] As a further method, the method of performing four-property inspection on the electronic archives and classifying the electronic archives into high-quality archives, qualified archives and problematic archives according to the inspection results includes: Obtain electronic archive metadata, verify the digital signature of the electronic archive, quantify the results of the signature verification algorithm, and obtain the authenticity index value; Obtain the number of bytes and hash value of the electronic file, and compare them with the standard number of bytes and standard hash value. If the hash matches and there is no byte difference, the integrity index value is 100. If the hash matches but there is a byte difference, the difference ratio is used as the index value. If the hash value does not match, the index value is 0 to obtain the integrity index value. Obtain the electronic archive format information, check whether it meets the specified format standards, quantify the error level, and obtain the usability index value; Quantify security based on the encryption strength of electronic archives and obtain security index values; Determine the weight of each indicator, the expression is: in, For the The weight of an indicator is Time represents authenticity, Represents integrity, Represents availability, Represents safety, and Respectively and The entropy value of an indicator, and Respectively and The coefficient of variation of each indicator; The comprehensive quality score of the electronic files is calculated by weighted summation, and the electronic files are divided into high-quality files, qualified files and problematic files according to the preset score range.

[0008] As a further method, the method of learning the file features of various electronic archives and allocating data streams to various electronic archives includes: Extract lexical features, semantic features and structural features, and count the feature types, feature weights and feature complexity of various electronic archives; Quantify the data flow size in bytes / unit time and calculate the data flow size allocated to each type of electronic archive. The expression is: in, Indicates The data stream size of the electronic archive, is the basic data stream reference value, For the The number of feature types in the class electronic archive, For the Electronic archives of the category The importance weight of the features, For the Electronic archives of the category The average complexity of the features, and They are respectively within the observation period, Class and The total number of times the electronic archives of this category have been accessed, is the average data interaction volume for a single access operation, For the The risk quantification value of electronic archives, For the The total number of files in the electronic archives of this category, is the number of all files; Bandwidth restrictions are imposed on various electronic files based on the size of the data stream, and the electronic files are compressed before data stream transmission.

[0009] As a further method, the method of analyzing the user's access behavior to various archive data streams and predicting the next archive accessed by the user based on the analysis results includes: Obtain user access operations on various electronic archive data streams, including access time, access duration, operation type, category and feature information of the archives accessed, as user access behavior logs; Through recurrent neural networks, the temporal patterns and logical relationships of user access behavior logs are learned, and the temporal patterns and logical relationships of user access behaviors are deeply modeled and learned to predict the next file that the user will access based on the current access information.

[0010] As a further method, the method of dynamically adjusting the user's access rights by using reinforcement learning according to the prediction results includes: Determine the user state space: Integrate the user's attributes, the user's currently accessed archive sequence information, the predicted next archive, and the user's historical access frequency vector into a state vector; Determine the user's action space, including maintaining current access rights, increasing access rights by one level, decreasing access rights by one level, and temporarily freezing access rights; The reward function is designed based on whether the user's access behavior meets business requirements and security policies. The expression is: in, For in time The user state space, For in time The user action space, For users In time The frequency of visits, is the maximum access frequency among all users, For users In time business progress, is the maximum business progress among all users, is the time attenuation coefficient, For users In time The request response time, For users In time The importance level of the archive being accessed, and are the coefficients for balancing the impact of business needs and security strategies on rewards, and , For the system at time Safety status indicators, For users In time Behavioral compliance scores, The difference between this permission adjustment action and the previous permission adjustment action. The coefficient for adjusting the amplitude of authority; In each time step, the user's permissions are dynamically adjusted according to the user's current state space and the feedback of the reward function to maximize the long-term cumulative rewards, and a reinforcement learning model is constructed to learn the permission adjustment strategy; After the training is completed, in actual operation, when the user performs an access operation, the system obtains the current state and selects the optimal action, that is, determines the access permission adjustment strategy and executes it.

[0011] As a further method, the method for constructing a knowledge graph based on the association of the semantic features of the electronic archive and the access rights of the user includes: Conduct semantic analysis on electronic archives to extract entities and relationships in electronic archives, including entity extraction, relationship extraction and attribute extraction; Determine the electronic archive entities and relationships that the user can access based on the user's access rights, integrate the extracted entities and relationships as well as the user's access rights, and construct a knowledge graph.

[0012] As a further method, the method for predicting the scope of risk propagation based on the location, connectivity and information propagation path of the nodes in the knowledge graph includes: By calculating the centrality index of each node in the knowledge graph, including degree centrality, closeness centrality and betweenness centrality, the risk transfer probability of the node is calculated, and the expression is: in, Representation Node To Node The probability of risk transfer between , and are the weight coefficients of degree centrality, closeness centrality and betweenness centrality, respectively. and Node and nodes The adjacency of is the total number of nodes, is a node set, and Node To Node and nodes To Node The shortest byte length, and Passing nodes and passing nodes Node To Node The shortest number of bytes, For Node To Node The shortest total number of bytes, For Node The set of adjacent nodes of is a constant, For Node and nodes The strength of semantic association between The risk transfer probability matrix of each archive relationship is weighted and summed based on the node connectivity weight, where the iterative expression of the node connectivity weight is: in, For Node The connection weight, is the damping coefficient, To point to a node The set of all nodes of For Node The connection weight of An adjacency list is established based on the weighted sum matrix, and the risk propagation path is queried in the adjacency table to predict the scope of risk propagation.

[0013] As a further method, the method for temporarily isolating and backing up the affected electronic archives within the scope includes: Conduct risk level assessment on electronic files within the risk transmission scope and determine the priority of isolation and backup; For electronic archives with high priority, the isolation mechanism is automatically triggered and they are temporarily moved from the main storage system to the isolation area; Execute backup strategies, copy isolated electronic files to a secure backup storage system, and perform integrity checks on backup data to ensure that the data has not been tampered with or damaged during the backup process; Once the risk is eliminated, execute the electronic archive recovery process, including verifying the integrity of the backup data, restoring the data from the backup center, and updating the archive management system.

[0014] As a further method, the method of storing electronic files on the blockchain based on blockchain technology and recording each access and modification operation of the electronic files on the blockchain includes: The node identities in the blockchain are divided into three types: demander, storage party, and no storage relationship. The demander and the storage party sign a storage service smart contract, and the contract is put on the chain to take effect; The demander processes the electronic archive into storage blocks, publishes the storage demand information to the storage party, triggers the storage service smart contract, and the storage party stores the relevant storage blocks in its own storage space after verification, and generates storage transaction information and submits it to the blockchain; Every access and modification operation of the electronic file is recorded as transaction information through a smart contract and submitted to the blockchain. The demander and the storage party manage the storage block based on the storage block information and storage transaction information.

[0015] The second aspect of the present invention provides a standard electronic archive management system based on artificial intelligence, comprising: The archive four-property inspection module is used to inspect the four properties of electronic archives, classify the electronic archives into high-quality archives, qualified archives and problem archives according to the inspection results, learn the file characteristics of various electronic archives, and allocate data streams for various electronic archives; the four-property inspection includes authenticity, integrity, availability and security inspection; The access rights control module is used to analyze the user's access behavior to various archive data streams, predict the next archive that the user will access based on the analysis results, and dynamically adjust the user's access rights using reinforcement learning based on the prediction results; A knowledge graph construction module, used to construct a knowledge graph based on the association of semantic features of the electronic archive and the access rights of the user; A risk range prediction module, used to predict the risk propagation range according to the location, connectivity and information propagation path of the nodes in the knowledge graph, and to temporarily isolate and back up the electronic files affected within the range; The information chain storage module is used to store electronic files on the chain based on blockchain technology, and record every access and modification operation of the electronic files on the blockchain to obtain a complete operation audit chain.

[0016] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) Through in-depth analysis of user behavior and prediction of the next archive access, the present invention can optimize resource allocation in advance, reasonably allocate archive data flow resources, reduce unnecessary network bandwidth occupation and data transmission delay, and thus significantly improve the convenience and efficiency of users in obtaining archive information; (2) The present invention dynamically adjusts user access rights. When the system predicts that a user needs to access a certain file, it automatically temporarily increases the user's rights, thereby reducing the complexity of the permission application process. (3) The present invention ensures transparent operations, secure and tamper-proof data through the use of blockchain, maintains the authority and credibility of archive management, and improves the overall intelligence, efficiency and security of electronic archive management. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flowchart of a standard electronic archive management method based on artificial intelligence in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0019] Reference Figure 1 As shown, the present invention provides a standard electronic archive management method based on artificial intelligence, comprising: S100 performs four-property inspection on electronic archives, classifies electronic archives into high-quality archives, qualified archives and problematic archives according to the inspection results, learns the file characteristics of various electronic archives, and allocates data streams for various electronic archives; the four-property inspection includes authenticity, integrity, availability and security inspection; In the actual evaluation, taking the electronic contract file of a certain enterprise as an example, after obtaining its metadata, the result of the signature verification algorithm shows that the authenticity index value of the contract reaches 90; the number of bytes of the obtained electronic contract file is 500KB, and the hash value matches the standard hash value, but there is a difference of 100B compared with the standard number of bytes. According to the formula, the integrity index value is 99.8%; the format of the electronic contract file is PDF, which fully complies with the format standard specified by the enterprise, and the availability index value is 100; the AES-256 encryption algorithm is used, and the security index value is 80 after security quantification; the weights of each indicator are: authenticity weight 0.3, integrity weight 0.3, availability weight 0.2, security weight 0.2, and the weighted sum calculation of the comprehensive quality score is 92.94. According to the preset score range, the electronic contract file is classified as a high-quality file; In the actual evaluation, there are 10 types of statistical features, the average feature weight is 0.5, the average feature complexity is 3, the basic data flow reference value is 1000, the total number of electronic contract archives accessed during the observation period is 50, and the average data interaction volume of simple access operations is 8KB. There are 287 high-quality archives and the total number of archives is 500. The calculated data flow size is 829444 bytes / 1 minute.

[0020] S200 analyzes the user's access behavior to various archive data streams, predicts the next archive the user will access based on the analysis results, and dynamically adjusts the user's access rights using reinforcement learning based on the prediction results; In the actual assessment, the user's access operation to the electronic contract archive data stream is recorded. A user accesses the contract archive at 10:00 on October 1, 2024, the access duration is 30 minutes, the operation type is viewing, the category of the archive accessed is contract, and the characteristic information is information such as specific terms as the user access behavior log; these logs are learned by recurrent neural networks. After a large amount of data training, it is possible to predict the next archive that the user may access based on the current access information, and predict that the next archive that may be accessed is the related supplementary agreement archive of the same project; In the actual evaluation, the user state space is determined: a user has employee attributes, the current sequence information of the accessed electronic contract files is [Contract 1, Contract 2], the next file is predicted to be Contract 3, and the historical access frequency vector is [0.2, 0.3], which is integrated into the state vector; the action space is determined: including maintaining the current access rights, improving the first level of access rights, reducing the first level of access rights, and temporarily freezing the access rights; the reward value is calculated through the reward function, and the user's permissions are dynamically adjusted according to the user's current state space and the feedback of the reward function, specifically improving the first level of access rights.

[0021] S300 builds a knowledge graph based on the association of semantic features of the electronic archive and the access rights of the user; In the actual assessment, semantic analysis is performed on the electronic contract files to extract entities such as Party A Company, Party B Company, contract amount, and other relationships such as signing relationship and inclusion relationship, and attribute extraction is performed including the value of the contract amount, signing date, etc.; the entities and relationships accessible to the user are determined based on user access rights, among which ordinary employees can only access contract entities and relationships related to their own business, and this information is integrated to build a knowledge graph.

[0022] S400 predicts the scope of risk propagation based on the location, connectivity and information propagation path of the nodes in the knowledge graph, and temporarily isolates and backs up the affected electronic files within the scope; In the actual evaluation, the centrality index of the electronic contract archive node in the knowledge graph is calculated. For node A (electronic contract archive) and node B (related supplementary agreement archive), the total number of nodes is 94, the shortest number of bytes passing through nodes A and B are 20 and 15 respectively, the total number of shortest bytes is 50, the adjacent node sets are 5 and 3 nodes respectively, the semantic association strength is 0.6, and the risk transfer probability from node A to node B is 0.623 according to the risk transfer probability formula; In the actual assessment, the risk transfer probability matrix of each archive relationship is weighted and summed based on the node connectivity weight. The weight is calculated according to the node connectivity weight iteration formula, where the damping coefficient is 0.85. The adjacency table is established based on the weighted sum matrix, the risk propagation path is queried, and the scope of risk propagation is predicted. It is predicted that if a key contract file has a risk, it will spread to the 5 contract files that have direct business associations with it. In the actual assessment, the five contract files were assessed for risk level, and the priority of isolation and backup was determined. According to the priority order, they were temporarily moved from the main storage system to the isolation area, which uses an independent storage network and is isolated from the external network. The backup strategy was implemented to copy the isolated electronic contract files to a safe backup storage system, which uses redundant storage technology RAID5 array. The backup data was also checked for integrity, and the hash value was calculated for comparison to ensure that the data was not tampered with or damaged during the backup process.

[0023] S500 stores electronic files on the blockchain based on blockchain technology, and records every access and modification operation of the electronic files on the blockchain to obtain a complete operation audit chain.

[0024] In the actual evaluation, the node identities in the blockchain are divided into three types: demanders (various departments within the enterprise), storage parties (professional cloud storage service providers) and those with no storage relationship. The demanders process the electronic contract files into storage blocks, each of which is 1MB in size, and publish the storage demand information to the storage party, triggering the storage service smart contract. After verification, the storage party stores the relevant storage blocks in its own storage space and generates storage transaction information and submits it to the blockchain.

[0025] In this embodiment, the method of performing four-property inspection on electronic files and classifying the electronic files into high-quality files, qualified files and problem files according to the inspection results includes: Obtain electronic archive metadata, verify the digital signature of the electronic archive, quantify the results of the signature verification algorithm, and obtain the authenticity index value; Obtain the number of bytes and hash value of the electronic file, and compare them with the standard number of bytes and standard hash value. If the hash matches and there is no byte difference, the integrity index value is 100. If the hash matches but there is a byte difference, the difference ratio is used as the index value. If the hash value does not match, the index value is 0 to obtain the integrity index value. Obtain the electronic archive format information, check whether it meets the specified format standards, quantify the error level, and obtain the usability index value; Quantify security based on the encryption strength of electronic archives and obtain security index values; Determine the weight of each indicator, the expression is: in, For the The weight of an indicator is Time represents authenticity, Represents integrity, Represents availability, Represents safety, and Respectively and The entropy value of an indicator, and Respectively and The coefficient of variation of each indicator; The comprehensive quality score of the electronic files is calculated by weighted summation, and the electronic files are divided into high-quality files, qualified files and problematic files according to the preset score range.

[0026] In this embodiment, the method of learning the file features of various electronic archives and allocating data streams to various electronic archives includes: Extract lexical features, semantic features and structural features, and count the feature types, feature weights and feature complexity of various electronic archives; Quantify the data flow size in bytes / unit time and calculate the data flow size allocated to each type of electronic archive. The expression is: in, Indicates The data stream size of the electronic archive, is the basic data stream reference value, For the The number of feature types in the class electronic archive, For the Electronic archives of the category The importance weight of the features, For the Electronic archives of the category The average complexity of the features, and They are respectively within the observation period, Class and The total number of times the electronic archives of this category have been accessed, is the average data interaction volume for a single access operation, For the The risk quantification value of electronic archives, For the The total number of files in the electronic archives of this category, is the number of all files; Bandwidth restrictions are imposed on various electronic files based on the size of the data stream, and the electronic files are compressed before data stream transmission.

[0027] In this embodiment, the method of analyzing the user's access behavior to various archive data streams and predicting the next archive accessed by the user according to the analysis result includes: Obtain user access operations on various electronic archive data streams, including access time, access duration, operation type, category and feature information of the archives accessed, as user access behavior logs; Through recurrent neural networks, the temporal patterns and logical relationships of user access behavior logs are learned, and the temporal patterns and logical relationships of user access behaviors are deeply modeled and learned to predict the next file that the user will access based on the current access information.

[0028] In this embodiment, the method of dynamically adjusting the user's access rights by using reinforcement learning according to the prediction results includes: Determine the user state space: Integrate the user's attributes, the user's currently accessed archive sequence information, the predicted next archive, and the user's historical access frequency vector into a state vector; Determine the user's action space, including maintaining current access rights, increasing access rights by one level, decreasing access rights by one level, and temporarily freezing access rights; The reward function is designed based on whether the user's access behavior meets business requirements and security policies. The expression is: in, For in time The user state space, For in time The user action space, For users In time The frequency of visits, is the maximum access frequency among all users, For users In time business progress, is the maximum business progress among all users, is the time attenuation coefficient, For users In time The request response time, For users In time The importance level of the archive being accessed, and are the coefficients for balancing the impact of business needs and security strategies on rewards, and , For the system at time Safety status indicators, For users In time Behavioral compliance scores, The difference between this permission adjustment action and the previous permission adjustment action. The coefficient for adjusting the amplitude of authority; In each time step, the user's permissions are dynamically adjusted according to the user's current state space and the feedback of the reward function to maximize the long-term cumulative rewards, and a reinforcement learning model is constructed to learn the permission adjustment strategy; After the training is completed, in actual operation, when the user performs an access operation, the system obtains the current state and selects the optimal action, that is, determines the access permission adjustment strategy and executes it.

[0029] In this embodiment, the method for constructing a knowledge graph based on the association of the semantic features of the electronic archive and the access rights of the user includes: Conduct semantic analysis on electronic archives to extract entities and relationships in electronic archives, including entity extraction, relationship extraction and attribute extraction; Determine the electronic archive entities and relationships that the user can access based on the user's access rights, integrate the extracted entities and relationships as well as the user's access rights, and construct a knowledge graph.

[0030] In this embodiment, the method for predicting the scope of risk propagation based on the location, connectivity and information propagation path of the nodes in the knowledge graph includes: By calculating the centrality index of each node in the knowledge graph, including degree centrality, closeness centrality and betweenness centrality, the risk transfer probability of the node is calculated, and the expression is: in, Representation Node To Node The probability of risk transfer between , and are the weight coefficients of degree centrality, closeness centrality and betweenness centrality, respectively. and Node and nodes The adjacency of is the total number of nodes, is a node set, and Node To Node and nodes To Node The shortest byte length, and Passing nodes and passing nodes Node To Node The shortest number of bytes, For Node To Node The shortest total number of bytes, For Node The set of adjacent nodes of is a constant, For Node and nodes The strength of semantic association between The risk transfer probability matrix of each archive relationship is weighted and summed based on the node connectivity weight, where the iterative expression of the node connectivity weight is: in, For Node The connection weight, is the damping coefficient, To point to a node The set of all nodes of For Node The connection weight of An adjacency list is established based on the weighted sum matrix, and the risk propagation path is queried in the adjacency table to predict the scope of risk propagation.

[0031] In this embodiment, the method for temporarily isolating and backing up the affected electronic files within the scope includes: Conduct risk level assessment on electronic files within the risk transmission scope and determine the priority of isolation and backup; For electronic archives with high priority, the isolation mechanism is automatically triggered and they are temporarily moved from the main storage system to the isolation area; Execute backup strategies, copy isolated electronic files to a secure backup storage system, and perform integrity checks on backup data to ensure that the data has not been tampered with or damaged during the backup process; Once the risk is eliminated, execute the electronic archive recovery process, including verifying the integrity of the backup data, restoring the data from the backup center, and updating the archive management system.

[0032] In this embodiment, the method of storing electronic files on the blockchain based on blockchain technology and recording each access and modification operation of the electronic files on the blockchain includes: The node identities in the blockchain are divided into three types: demander, storage party, and no storage relationship. The demander and the storage party sign a storage service smart contract, and the contract is put on the chain to take effect; The demander processes the electronic archive into storage blocks, publishes the storage demand information to the storage party, triggers the storage service smart contract, and the storage party stores the relevant storage blocks in its own storage space after verification, and generates storage transaction information and submits it to the blockchain; Every access and modification operation of the electronic file is recorded as transaction information through a smart contract and submitted to the blockchain. The demander and the storage party manage the storage block based on the storage block information and storage transaction information.

[0033] The second aspect of the present invention provides a standard electronic archive management system based on artificial intelligence, comprising: The archive four-property inspection module is used to inspect the four properties of electronic archives, classify the electronic archives into high-quality archives, qualified archives and problem archives according to the inspection results, learn the file characteristics of various electronic archives, and allocate data streams for various electronic archives; the four-property inspection includes authenticity, integrity, availability and security inspection; The access rights control module is used to analyze the user's access behavior to various archive data streams, predict the next archive that the user will access based on the analysis results, and dynamically adjust the user's access rights using reinforcement learning based on the prediction results; A knowledge graph construction module, used to construct a knowledge graph based on the association of semantic features of the electronic archive and the access rights of the user; A risk range prediction module, used to predict the risk propagation range according to the location, connectivity and information propagation path of the nodes in the knowledge graph, and to temporarily isolate and back up the electronic files affected within the range; The information chain storage module is used to store electronic files on the chain based on blockchain technology, and record every access and modification operation of the electronic files on the blockchain to obtain a complete operation audit chain.

[0034] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A standard electronic archive management method based on artificial intelligence, characterized in that: The following steps are involved: Conduct four-property inspections on electronic archives, classify them into high-quality archives, qualified archives, and problematic archives based on the inspection results, learn the file characteristics of various electronic archives, and assign data streams to various electronic archives; The four-property inspection includes authenticity, integrity, availability and security inspection; Analyze the user's access behavior to various archive data streams, predict the next archive the user will access based on the analysis results, and dynamically adjust the user's access rights using reinforcement learning based on the prediction results; Building a knowledge graph based on the association of semantic features of electronic archives and the access rights of the users; According to the location, connectivity and information propagation path of the nodes in the knowledge graph, the scope of risk propagation is predicted, and the affected electronic files within the scope are temporarily isolated and backed up; Based on blockchain technology, electronic files are stored on the chain, and every access and modification operation of the electronic files is recorded on the blockchain to obtain a complete operation audit chain.

2. According to claim 1, a standard electronic archive management method based on artificial intelligence is characterized in that: The method of performing four-property inspection on electronic archives and classifying the electronic archives into high-quality archives, qualified archives and problematic archives according to the inspection results includes: Obtain electronic archive metadata, verify the digital signature of the electronic archive, quantify the results of the signature verification algorithm, and obtain the authenticity index value; Obtain the number of bytes and hash value of the electronic file, and compare them with the standard number of bytes and standard hash value. If the hash matches and there is no byte difference, the integrity index value is 100. If the hash matches but there is a byte difference, the difference ratio is used as the index value. If the hash value does not match, the index value is 0 to obtain the integrity index value. Obtain the electronic archive format information, check whether it meets the specified format standards, quantify the error level, and obtain the usability index value; Quantify security based on the encryption strength of electronic archives and obtain security index values; Determine the weight of each indicator, the expression is: in, For the The weight of an indicator is Time represents authenticity, Represents integrity, Represents availability, Represents safety, and Respectively and The entropy value of an indicator, and Respectively and The coefficient of variation of each indicator; The comprehensive quality score of the electronic files is calculated by weighted summation, and the electronic files are divided into high-quality files, qualified files and problematic files according to the preset score range.

3. The standard electronic archive management method based on artificial intelligence according to claim 1 is characterized in that: The method of learning the file characteristics of various electronic archives and allocating data streams to various electronic archives includes: Extract lexical features, semantic features and structural features, and count the feature types, feature weights and feature complexity of various electronic archives; Quantify the data flow size in bytes / unit time and calculate the data flow size allocated to each type of electronic archive. The expression is: in, Indicates The data stream size of the electronic archive, is the basic data stream reference value, For the The number of feature types in the class electronic archive, For the Electronic archives of the category The importance weight of the features, For the Electronic archives of the category The average complexity of the features, and They are respectively within the observation period, Class and The total number of times the electronic archives of this category have been accessed, is the average data interaction volume for a single access operation, For the The risk quantification value of electronic archives, For the The total number of files in the electronic archives of this category, is the number of all files; Bandwidth restrictions are imposed on various electronic files based on the size of the data stream, and the electronic files are compressed before data stream transmission.

4. The standard electronic archive management method based on artificial intelligence according to claim 1, characterized in that: The method of analyzing the user's access behavior to various archive data streams and predicting the next archive to be accessed by the user according to the analysis result includes: Obtain user access operations on various electronic archive data streams, including access time, access duration, operation type, category and feature information of the archives accessed, as user access behavior logs; Through recurrent neural networks, the temporal patterns and logical relationships of user access behavior logs are learned, and the temporal patterns and logical relationships of user access behaviors are deeply modeled and learned to predict the next file that the user will access based on the current access information.

5. The standard electronic archive management method based on artificial intelligence according to claim 1 is characterized in that: The method for dynamically adjusting the user's access rights by using reinforcement learning according to the prediction results includes: Determine the user state space: Integrate the user's attributes, the user's currently accessed archive sequence information, the predicted next archive, and the user's historical access frequency vector into a state vector; Determine the user's action space, including maintaining current access rights, increasing access rights by one level, decreasing access rights by one level, and temporarily freezing access rights; The reward function is designed based on whether the user's access behavior meets business requirements and security policies. The expression is: in, For in time The user state space, For in time The user action space, For users In time The frequency of visits, is the maximum access frequency among all users, For users In time business progress, is the maximum business progress among all users, is the time attenuation coefficient, For users In time The request response time, For users In time The importance level of the archive being accessed, and are the coefficients for balancing the impact of business needs and security strategies on rewards, and , For the system at time Safety status indicators, For users In time Behavioral compliance scores, The difference between this permission adjustment action and the previous permission adjustment action. The coefficient for adjusting the amplitude of authority; In each time step, the user's permissions are dynamically adjusted according to the user's current state space and the feedback of the reward function to maximize the long-term cumulative rewards, and a reinforcement learning model is constructed to learn the permission adjustment strategy; After the training is completed, in actual operation, when the user performs an access operation, the system obtains the current state and selects the optimal action, that is, determines the access permission adjustment strategy and executes it.

6. The standard electronic archive management method based on artificial intelligence according to claim 1 is characterized in that: The method for constructing a knowledge graph based on the association of electronic archive semantic features and the user's access rights includes: Conduct semantic analysis on electronic archives to extract entities and relationships in electronic archives, including entity extraction, relationship extraction and attribute extraction; Determine the electronic archive entities and relationships that the user can access based on the user's access rights, integrate the extracted entities and relationships as well as the user's access rights, and construct a knowledge graph.

7. The standard electronic archive management method based on artificial intelligence according to claim 1 is characterized in that: The method for predicting the scope of risk propagation based on the location, connectivity and information propagation path of the nodes in the knowledge graph includes: By calculating the centrality index of each node in the knowledge graph, including degree centrality, closeness centrality and betweenness centrality, the risk transfer probability of the node is calculated, and the expression is: in, Representation Node To Node The probability of risk transfer between , and are the weight coefficients of degree centrality, closeness centrality and betweenness centrality, respectively. and Node and nodes The adjacency of is the total number of nodes, is a node set, and Node To Node and nodes To Node The shortest byte length, and Passing nodes and passing nodes Node To Node The shortest number of bytes, For Node To Node The shortest total number of bytes, For Node The set of adjacent nodes of is a constant, For Node and nodes The strength of semantic association between The risk transfer probability matrix of each archive relationship is weighted and summed based on the node connectivity weight, where the iterative expression of the node connectivity weight is: in, For Node The connection weight, is the damping coefficient, To point to a node The set of all nodes of For Node The connection weight of An adjacency list is established based on the weighted sum matrix, and the risk propagation path is queried in the adjacency table to predict the scope of risk propagation.

8. The standard electronic archive management method based on artificial intelligence according to claim 1 is characterized in that: The method for temporarily isolating and backing up the affected electronic archives within the scope includes: Conduct risk level assessment on electronic files within the risk transmission scope and determine the priority of isolation and backup; For electronic archives with high priority, the isolation mechanism is automatically triggered and they are temporarily moved from the main storage system to the isolation area; Execute backup strategies, copy isolated electronic files to a secure backup storage system, and perform integrity checks on backup data to ensure that the data has not been tampered with or damaged during the backup process; Once the risk is eliminated, execute the electronic archive recovery process, including verifying the integrity of the backup data, restoring the data from the backup center, and updating the archive management system.

9. The standard electronic archive management method based on artificial intelligence according to claim 1, characterized in that: The method of storing electronic files on the blockchain based on blockchain technology and recording each access and modification operation of the electronic files on the blockchain includes: The node identities in the blockchain are divided into three types: demander, storage party, and no storage relationship. The demander and the storage party sign a storage service smart contract, and the contract is put on the chain to take effect; The demander processes the electronic archive into storage blocks, publishes the storage demand information to the storage party, triggers the storage service smart contract, and the storage party stores the relevant storage blocks in its own storage space after verification, and generates storage transaction information and submits it to the blockchain; Every access and modification operation of the electronic file is recorded as transaction information through a smart contract and submitted to the blockchain. The demander and the storage party manage the storage block based on the storage block information and storage transaction information.

10. A standard electronic archive management system based on artificial intelligence, used to implement a standard electronic archive management method based on artificial intelligence as claimed in any one of claims 1 to 9, characterized in that: The system comprises: The archive four-property inspection module is used to inspect the four properties of electronic archives, classify the electronic archives into high-quality archives, qualified archives and problem archives according to the inspection results, learn the file characteristics of various electronic archives, and allocate data streams for various electronic archives; the four-property inspection includes authenticity, integrity, availability and security inspection; The access rights control module is used to analyze the user's access behavior to various archive data streams, predict the next archive that the user will access based on the analysis results, and dynamically adjust the user's access rights using reinforcement learning based on the prediction results; A knowledge graph construction module, used to construct a knowledge graph based on the association of semantic features of the electronic archive and the access rights of the user; A risk range prediction module, used to predict the risk propagation range according to the location, connectivity and information propagation path of the nodes in the knowledge graph, and to temporarily isolate and back up the electronic files affected within the range; The information chain storage module is used to store electronic files on the chain based on blockchain technology, and record every access and modification operation of the electronic files on the blockchain to obtain a complete operation audit chain.

Images