A verifiable control method and system for employee subject authorization verification and authorization change

Abstract

This invention discloses a verifiable control method and system for employee subject authorization verification and authorization changes. The method segments the natural language authorization task and establishes session isolation for each subtask; in the first session, a neural network model is invoked to output a fixed-format set of employee subject names, action enumerations, and task phrase sets; a candidate task subset is generated based on cross-indexing, and in the second session, a neural network model is invoked to perform constrained semantic alignment, allowing the output of candidate task codes or newly created task names; the execution layer rejects out-of-bounds output through a hard check gate; a compensation transaction log is generated before writing and a snapshot and fingerprint value of the affected domain file before the change are saved; after writing, the fingerprint value is recorded and a commit flag is written; in case of abnormal restart, rollback recovery can be performed based on the transaction status; the authorization cache is maintained or rebuilt according to the incremental and differential sets, providing reversal or compensation transactions based on fingerprint value conflict detection.

Description

Technical Field

[0001] This invention relates to a verifiable control method and system for employee authorization verification and authorization change, belonging to the field of information management data security control. Background Technology

[0002] In scenarios such as organizational management, park operations, energy and power, and manufacturing and maintenance, it is necessary to establish an authorization relationship between the "subject (personnel) and the authorized object (project / equipment / area / process)," and conduct on-site verification through QR codes, viewing codes, or near-field communication. In recent years, various technologies to improve the efficiency of information entry have developed rapidly, leading to increased efficiency. Among these, neural network model technology has developed rapidly, and parsing natural language tasks can lower the input threshold. However, there are uncontrollable factors regarding accuracy, especially when the size of the authorization pair set increases, names are similar, and abbreviations / synonyms exist, introducing uncertainties such as output out-of-bounds errors, format drift, semantic crosstalk, and fictitious objects. Under lightweight deployment conditions (file-based storage, no database transactions), these risks can easily evolve into problems such as inconsistencies between indexes and caches, difficulty in tracing incorrect authorizations, and difficulty in recovering from erroneous operations. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a verifiable control method and system for employee principal authorization verification and authorization changes, enabling low-cost, lightweight deployment and efficient, accurate verification control of authorization.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] This invention provides a verifiable control method for employee authorization verification and authorization change.

[0006] Includes the following steps,

[0007] 1) The data storage device stores employee details, task lists, and corresponding cross-index mapping tables;

[0008] 2) Employees input natural language commands through the login terminal, and the natural language commands are divided into multiple sub-tasks;

[0009] 3) Each subtask is parsed using a neural network model to obtain a list of employee main names, action enumeration commands, and task names. The action enumeration commands include at least: add authorization, revoke authorization, set authorization list, clear authorization list, and query only. The list of employee main names, action enumeration commands, and task names have a fixed format, and the fixed format conforms to the limited field types, the maximum number of fields, and the maximum string length.

[0010] 4) Read the cross-index mapping table based on the task name, generate a list of candidate tasks, and select the top N candidate tasks by score, where N is a fixed upper limit. The score is based on at least one or more of the following features: string similarity of task name, exact matching degree of task code, synonym / abbreviation mapping matching degree, and vectorized semantic similarity.

[0011] 5) Call the neural network model to transform the subtask into a regularized task. The regularized task is a list of tasks in the top N candidate task list or a newly created task list, where N is a fixed upper limit. The task list output by the regularized task must be in structured text or structured key-value pair text format, and must contain at least a task list field and an optional newly created task name field.

[0012] 6) Perform validation on the above task list. Validation includes employee name validation, action enumeration command validation, task name validation, task list existence validation, and output format consistency validation. The task list validation requires that the output task list does not exceed the candidate task list.

[0013] 7) For the task list that meets the verification specifications, generate an associated index with the employee subject / task that created the task, and lock the data based on the newly added task list, saving the data before and after the change;

[0014] 8) Perform idempotent authorization changes and updates on the data stored in the data storage. Idempotent authorization includes enumerating the command change task list by action. When adding authorization, the change task list is obtained by merging the original task list and the task change list. When revoking authorization, the change task list is obtained by removing the task change list from the original task list. When resetting the task list, the original task list is replaced with the task change list. When clearing the task list, the task change list is cleared.

[0015] 9) And update the cross-index mapping table synchronously to make the login terminal consistent with the data storage.

[0016] According to the verifiable control method for employee main authorization verification and authorization change, in step 2), first check whether there is a predetermined batch processing separator mark; if there is, perform segmentation processing according to step 2); if there is not, treat the entire natural language command as a single subtask. The separator mark is a preset or configured segmentation symbol, sequence or newline.

[0017] According to the verifiable control method for employee main authorization verification and authorization change, in step 3), the intent of each subtask is parsed and cross-indexed and mapped, and each subtask does not share the state with each other.

[0018] According to the verifiable control method for employee subject authorization verification and authorization change, in step 6), the employee subject name verification requires that the employee subject name be a substring of the original subtask text or a traceable evidence fragment; the task name verification requires that the output task names all belong to the task list of the top N candidate task list, and the newly created task list meets the creation rules; the task list existence verification requires that the task list existence verification be performed on the top N candidate task list, and a new task list is generated after the newly created task list is created according to the strategy; the output format consistency verification requires that the output meets the preset structured format, the upper limit of the number of lists, and the deduplication and sorting rules.

[0019] If any check fails, the task list will not be updated and the process will stop.

[0020] According to the verifiable control method for employee main authorization verification and authorization change, the newly created task list in step 5) must meet at least one or more creation rules: naming regular expression constraint, domain field integrity constraint, approval identifier existence constraint, and the initial state of the newly created task does not participate in task execution.

[0021] According to the verifiable control method for employee principal authorization verification and authorization change, in step 8), the data undergoes idempotent authorization change update verification. After passing the verification, a compensation transaction log needs to be recorded. The minimum set of files that needs to be written, maintained in the cache, and whose consistency verification needs to be performed is required. The compensation transaction log records at least: transaction number, transaction status, file identifier of the minimum set of files, fingerprint before change, fingerprint after change, and commit flag. The writing adopts the method of "writing temporary file → atomic renaming", and the commit flag is written to an independent flag file or log field after the commit.

[0022] According to the verifiable control method for employee subject authorization verification and authorization change, step 9) also includes authorization cache maintenance: for each task in the task list, the cross-index relationship between the task list and the employee subject is incrementally updated according to the addition of new tasks and removal of tasks; when a corresponding task is detected as missing, or the employee subject details are missing / damaged, or the consistency check fails, reconstruction is performed, and the employee subject and the task list are re-indexed and associated.

[0023] This invention provides a system with verifiable control methods for employee authorization verification and authorization changes.

[0024] It includes a login terminal, a processor, a data storage device, an authorization verification and authorization change module, and a neural network model. The login terminal is used for employee login and input, and displays work tasks. The data storage device stores the employee database, task list, and corresponding cross-index mapping table. The neural network model parses and outputs the natural language input from the employee. The authorization verification and authorization change module implements verifiable control methods.

[0025] This invention uses a neural network model to implement index constraints, verification, and idempotent authorization for employee login and task lists, enabling verifiable execution of authorization changes and maintaining consistency in the verification-control-alarm chain.

[0026] The advantage of this invention is that it can efficiently and accurately complete the above-mentioned authorization under low-cost and lightweight deployment conditions. Under the condition that the neural network model participates in the natural language authorization change, the execution layer has verifiability, controllability and recoverability, thereby avoiding out-of-bounds writing, and maintaining consistency, traceability and consistency recovery capabilities under large-scale task sets and lightweight storage conditions. Attached Figure Description

[0027] Figure 1 This is a flowchart of the method of the present invention;

[0028] Figure 2 Flowchart for enhancing matching and object standardization in constrained retrieval;

[0029] Figure 3 Flowchart for compensating for transaction log and consistency recovery process;

[0030] Figure 4 To verify the event-driven control and alarm linkage flowchart;

[0031] Figure 5 This is a diagram showing the relationship between cross-indexes, entity files, and authorization caches. Detailed Implementation

[0032] The following provides a preferred and complete embodiment of the present invention, describing the entire replicable process from input to output. (See also: [link to original text]) Figures 1-5 .

[0033] The system hardware of this invention includes a login terminal, a processor, a data storage device, an authorization verification and authorization change module, and a neural network model. The login terminal is used for employee login and input, and displays work tasks. The data storage device stores an employee database, a task list, and a corresponding cross-index mapping table. The neural network model parses and outputs the natural language input by the employee. The authorization verification and authorization change module implements verifiable control methods.

[0034] The data storage uses file-based entity records and cross-indexes to organize data. The employee subject index maps employee subject names to unique employee subject identifiers; the task index maps task names / task codes to task details; the employee subject detail file contains at least employee subject attributes and a list of authorized task content sets and authorized task content lists; the task list contains at least task attributes and a list of authorized employee subject identifiers with authorized cache fields.

[0035] The list of authorized employee identifiers is renewable data, a redundant cached field in the task details. Its content is calculated based on the list of authorized task content in each employee's subject details, according to defined rules.

[0036] For any task code c, iterate through the employee subject index and read the list of authorized task content for the corresponding employee subject details. Take the employee subject identifiers that satisfy "c∈Authorized Task Content List", form a set U_c, sort and remove duplicates, and write it back as the list of authorized employee subject identifiers.

[0037] Therefore, when this field is missing, corrupted, or fails consistency verification, it can be reconstructed using the above-described traversal and recalculation method without affecting the correctness of the authorization determination.

[0038] The data stored in the data storage is defined as follows: G_old represents the set of authorized task content of the employee subject before the authorization change; G_new represents the set of authorized task content after the change, obtained by enumerating the actions and operating on G_old and the target set T; T represents the set of authorized task content associated with the current instruction (or the set of task names to be created, which is determined by the constrained semantic alignment stage and generated by the program); Δ_add represents the newly added differential set, that is, the set of authorized task content added by G_new relative to G_old; Δ_del represents the deleted differential set, that is, the set of authorized task content removed by G_old relative to G_new; A represents the affected task set, which is the union of Δ_add and Δ_del, used to limit the scope of tasks that need to maintain cache, index, and transaction impact domain.

[0039] The verifiable control method for employee principal authorization verification and authorization change using the above system shall be executed according to steps S101 to S111:

[0040] S101 (Input Acquisition): The login terminal receives natural language authorization tasks from employees, or receives authorization correction requests triggered by verification events generated by the verification terminal. Verification events include scanning QR codes and conducting entry communications.

[0041] S102 (Task Segmentation): Detects whether a predefined batch processing separator exists; if it exists, it is split into multiple subtasks according to the separator and a processing queue is established; if it does not exist, the entire task is processed as a single subtask. The separator is a system-preset or configured segmentation symbol / sequence, or it can be a detectable form such as a newline segment.

[0042] S103 (Task Isolation Execution): Establish independent sessions and execution contexts for each subtask sequence to ensure that the parsing, mapping, and writing of each subtask do not share state.

[0043] S104 (Phase 1 Intent Parsing): In the first independent session, the neural network model is invoked to output fixed-format text, which includes at least: a set of employee subject names, an action enumeration (add authorization / revoke authorization / set authorization set / clear authorization set / query only), and a task phrase set. The execution layer of the neural network model parses this fixed-format text.

[0044] S105 (Candidate Retrieval and Truncation): A candidate list is obtained by performing a retrieval and scoring on the task phrase in the local task index; then, it is truncated into the top N candidate task subsets based on the scores, where N is a fixed upper limit. To avoid exceeding the prompt length limit, the number of candidates n that can be accommodated can be calculated based on the current prompt base length and the average length of candidate entries, while maintaining the upper limit of N, and the actual number of candidates is the smaller of N and n.

[0045] S106 (Phase 2 Constrained Semantic Alignment): In the second independent session, the neural network model is invoked, allowing only the output of the task code set within the candidate task subset, or the output of the new task name set; the execution layer rejects candidate task codes outside the candidate subset.

[0046] S107 (Hard Validation Gate): Performs employee subject source verification (the employee subject name must be a substring of the original subtask text or a traceable evidence fragment), action enumeration verification, task code existence verification, and output format / length / deduplication consistency verification; if any verification fails, writing is rejected and the reason is returned.

[0047] S108 (Task Creation and Identifier Generation): When an employee entity or task does not exist and the policy allows it, a unique identifier is generated according to the counter and an entity file is created, while the cross-index is updated; strong constraint rules or approval policies are applied to the "New Task Name Set", and creation is rejected if the conditions are not met.

[0048] S109 (Compensation Transaction Log Preparation): Generate transaction number and acquire write lock; construct the set of impact domain files for this change (index, counter, employee main details, task details and their cached fields) based on differential impact domain A; save the snapshot before the change and record the fingerprint value in the manner of "write temporary file → atomic rename"; set the transaction state to ready.

[0049] S110 (Idempotent Authorization Change and Consistency Maintenance): Performs idempotent updates on the authorized task content set of the employee entity, locates the affected task set based on the difference set, then updates the cross-index and maintains the authorization cache to ensure consistency between the verification end and the management end. Its implementation includes the following sub-steps:

[0050] S1101 (Set-based representation): The list of authorized task contents in the employee subject details is processed as a set, denoted as the old set G_old; the set of target task codes obtained in this standardization is denoted as T.

[0051] S1102 (Solve the new set by enumerating actions):

[0052] If the action is to add a new authorization, then G_new = G_old ∪ T; if the action is to revoke an authorization, then G_new = G_old ∪ T; if the action is to set the authorization set, then G_new = T; if the action is to clear the authorization set, then G_new = ∅ (empty set); if the action is to query only, then G_new = G_old and the write process is not initiated. The above definition ensures that repeated execution of the same action on the same input yields the same G_new, thus satisfying idempotency. During write-back, G_new is sorted, deduplicated, and serialized into a structured key-value pair text array, guaranteeing the uniqueness of the serialization result for the same set.

[0053] S1103 (Differential set location of affected tasks): Calculate Δ_add = G_new − G_old and Δ_del = G_old − G_new, and let the set of affected tasks A = Δ_add ∪ Δ_del. Specifically, A = Δ_add when adding authorization; A = Δ_del when revoking authorization; A is the union of the two when setting the authorization set; and A = G_old when clearing the authorization set.

[0054] S1104 (Write back employee subject details and synchronize index): Update the list of authorized tasks for the employee subject details to G_new, and update the name → employee subject identifier mapping in the employee subject index (add it if it is a newly created employee subject). The write operation uses "Write temporary file → Atomic rename".

[0055] S1105 (Cache Maintenance: Incremental Update and Reconstruction Switching): For each task code c∈A, read the list of authorized employee entity identifiers in the task details and process it according to the set U_c. If c∈Δ_add, add the employee entity identifier to U_c; if c∈Δ_del, remove the employee entity identifier from U_c; then sort U_c to remove duplicates and write it back. If a task detail is missing, the authorized employee entity identifier list field is missing / corrupted, or the consistency check fails after incremental update, switch to reconstruction mode: traverse the employee entity index and employee entity detail file, read the authorized task content list field in the employee entity detail for each employee entity identifier (this field represents the current authorized task content set for this employee entity); for the current task code c, collect all employee entity identifiers that satisfy "the authorized task content list contains c" to form a set U_c, sort U_c to remove duplicates, and write it back to the authorized employee entity identifier list field of the task details. The consistency check includes at least two-way verification: on the one hand, for any task code c, each employee entity identifier in the list of authorized employee entity identifiers should contain c in the list of authorized task content of the corresponding employee entity details; on the other hand, for any employee entity identifier, if its list of authorized task content contains a task code c, then the list of authorized employee entity identifiers in the task details should also contain that employee entity identifier; when either direction is not satisfied, the consistency check is determined to have failed and the above-mentioned reconstruction and write-back are performed.

[0056] S1106 (Consistency Check): Perform a two-way consistency check for each c∈A: the authorized task content list of the employee subject contains c if and only if the authorized employee subject identifier list of the task details contains the employee subject identifier; if not satisfied, enter the reconstruction mode and check again.

[0057] S111 (Commit, Output, and Consistency Recovery): Records the changed fingerprint and readable digest (including details of new / undoed / new / no changes) of the affected domain files, sets the transaction status to committed and writes a commit flag; outputs the process log; scans for incomplete transactions when the system restarts abnormally: performs rollback on transactions with transaction status = ready and missing commit flag, restores the snapshot before the change, and rebuilds the cache; during rollback, compares the current fingerprint value with the changed fingerprint file by file based on the transaction number to perform conflict detection. If they match, restores the snapshot before the change and rebuilds the cache; if they do not match, aborts safely to avoid overwriting subsequent manual modifications.

[0058] After the data server verification terminal of this invention generates a verification event, the system can use two equivalent paths for authorization determination: (a) determining whether the requested task belongs to the set of authorized task content lists for employee subjects; or (b) determining whether the employee subject identifier belongs to the set of authorized employee subject identifier lists for task details. If the determination passes, a release control signal is output or a work order task is generated; if the determination fails, an alarm event is generated and an alarm signal is output. Dynamic adjustment of rate limiting / freezing can be implemented using a sliding window counting mechanism: when the number of consecutive failures for the same employee subject within a time window W reaches a threshold K, a freeze record is written, and release is refused during the freeze period, while the frequency of alarm events is limited.

[0059] This invention can verify variations of the control method and further selection schemes.

[0060] (1) The candidate retrieval feature set can be expanded or pruned;

[0061] (2) For high-risk actions such as setting / clearing the authorization set, secondary confirmation or double-person review can be introduced;

[0062] (3) When the difference between candidate scores is lower than the threshold, resulting in insufficient disambiguation, the candidate list can be output and the writing process can be terminated, and manual confirmation can be requested.

[0063] (4) Reversal can be performed by “only allowing the most recent committed transaction to be reversed” or “generating a reverse compensation transaction” to maintain the consistent semantics of the transaction chain.

[0064] This invention enables authorized changes to have verifiable execution and consistency recovery capabilities through a collaborative closed loop of "candidate set constraints + hard check gates + idempotent execution + compensating transaction logs + minimum closure constraint of differential influence domain A + cache maintenance", while maintaining the consistency of the verification-control-alarm link.

[0065] Under low-cost and lightweight deployment conditions, the above authorization can be completed efficiently and accurately. Under the condition that the neural network model participates in the natural language authorization change, the execution layer has verifiability, controllability and recoverability, thereby avoiding out-of-bounds writing.

Claims

1. A method for verifiable control of employee authorization verification and authorization change, characterized in that, Includes the following steps, 1) The data storage device stores employee details, task lists, and corresponding cross-index mapping tables; 2) Employees input natural language commands through the login terminal, and the natural language commands are divided into multiple sub-tasks; 3) Each subtask is parsed using a neural network model to obtain a list of employee main names, action enumeration commands, and task names. The action enumeration commands include at least: add authorization, revoke authorization, set authorization list, clear authorization list, and query only. The list of employee main names, action enumeration commands, and task names have a fixed format, and the fixed format conforms to the limited field types, the maximum number of fields, and the maximum string length. 4) Read the cross-index mapping table based on the task name, generate a list of candidate tasks, and select the top N candidate tasks by score, where N is a fixed upper limit. The score is based on at least one or more of the following features: string similarity of task name, exact matching degree of task code, synonym / abbreviation mapping matching degree, and vectorized semantic similarity. 5) Call the neural network model to transform the subtask into a regularized task. The regularized task is a list of tasks in the top N candidate task list or a newly created task list, where N is a fixed upper limit. The task list output by the regularized task must be in structured text or structured key-value pair text format, and must contain at least a task list field and an optional newly created task name field. 6) Perform validation on the above task list. Validation includes employee name validation, action enumeration command validation, task name validation, task list existence validation, and output format consistency validation. The task list validation requires that the output task list does not exceed the candidate task list. 7) For the task list that meets the verification specifications, generate an associated index with the employee subject / task that created the task, and lock the data based on the newly added task list, saving the data before and after the change; 8) Perform idempotent authorization changes and updates on the data stored in the data storage. Idempotent authorization includes enumerating the command change task list by action. When adding authorization, the change task list is obtained by merging the original task list and the task change list. When revoking authorization, the change task list is obtained by removing the task change list from the original task list. When resetting the task list, the original task list is replaced with the task change list. When clearing the task list, the task change list is cleared. 9) And update the cross-index mapping table synchronously to make the login terminal consistent with the data storage.

2. The verifiable control method for employee principal authorization verification and authorization change according to claim 1, characterized in that, In step 2), first check if there is a predetermined batch processing delimiter; if there is, perform the segmentation process according to step 2); if there is no delimiter, treat the entire natural language command as a single subtask, where the delimiter is a preset or configured segmentation symbol, sequence, or newline.

3. The verifiable control method for employee principal authorization verification and authorization change according to claim 1, characterized in that, In step 3), each subtask's intent is parsed and cross-indexed in the mapping table, and each subtask does not share its state with others.

4. The verifiable control method for employee principal authorization verification and authorization change according to claim 1, characterized in that, In step 6), the employee subject name verification requires that the employee subject name be a substring of the original subtask text or a traceable evidence fragment. The task name verification requires that the output task names all belong to the task list of the top N candidate task list, and the newly created task list meets the creation rules. The task list existence verification requires that the task list existence verification be performed on the top N candidate task list, and a new task list is generated after the newly created task list is created according to the strategy. The output format consistency verification requires that the output meets the preset structured format, the upper limit of the number of lists, and the deduplication and sorting rules. If any check fails, the task list will not be updated and the process will stop.

5. The verifiable control method for employee principal authorization verification and authorization change according to claim 1, characterized in that, In step 5), the newly created task list must meet at least one or more creation rules: naming regular expression constraint, domain field integrity constraint, approval identifier existence constraint, and the initial state of the newly created task does not participate in task execution.

6. The verifiable control method for employee principal authorization verification and authorization change according to claim 1, characterized in that, In step 8), the data undergoes idempotent authorization change update verification. After passing the verification, it is necessary to record the compensation transaction log, write to it, maintain the cache, and perform consistency verification on the minimum set of files. The compensation transaction log shall record at least the following: transaction number, transaction status, file identifier of the minimum set of files, fingerprint before change, fingerprint after change, and commit flag; The write operation uses the "write temporary file → atomic rename" method, and the commit flag is written to a separate flag file or log field after the commit.

7. The verifiable control method for employee principal authorization verification and authorization change according to claim 1, characterized in that, Step 9) also includes authorized cache maintenance: for each task in the task list, incrementally update the cross-index relationship between the task list and the employee subject according to the addition and removal of tasks; when a corresponding task is missing, or the employee subject details are missing, damaged, or the consistency check fails, rebuild the system and retrieve the employee subject and re-index the task list.

8. A system for verifyable control of employee principal authorization verification and authorization change as described in any one of claims 1-7, characterized in that, It includes a login terminal, a processor, a data storage device, and a neural network model. The login terminal is used for employee login and input, and displays work tasks. The data storage device stores the employee database, task list, and corresponding cross-index mapping table. The neural network model parses and outputs the natural language input from the employee. The authorization verification and authorization change module implements verifiable control methods.

Images