Human resource data security management method and system

Abstract

The invention discloses a human resource data security management method and system, and particularly relates to the field of human resource data security management.The method comprises the steps that behavior log data are obtained, and a behavior event tetrad set containing behavior types, executor identities, execution contexts and timestamp fields is obtained after behavior dimension analysis is conducted on the behavior log data; performing combined recoding operation of system module marking, node marking and context nesting hierarchy marking on the behavior event tetrad set to obtain an initial behavior vector sequence; and performing time sequence window sliding and behavior type clustering operation on the initial behavior vector sequence to obtain continuous sub-behavior fragments serving as input sources of the multi-dimensional behavior sub-tensor. By constructing a behavior track tensor and a path redundancy projection space, a counterfeit behavior path which is hidden and injected in a cross-system consistency window period is identified, and structural level verification and tracing of a disguised legal behavior chain are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of human resource data security management. More specifically, the present invention relates to a human resource data security management method and system. Background Art

[0002] In large group enterprises, human resource management usually relies on multiple functionally heterogeneous system modules, such as organizational structure management, performance appraisal, salary payment, etc. The data exchange between these functional modules generally adopts an asynchronous replication or delayed synchronization mechanism, resulting in a short but exploitable consistency window period during the cross-system data transfer process;

[0003] During this period, if there are attackers or internal high-privilege users deliberately constructing operation paths, such as forging approval nodes, inserting abnormal performance records, or manipulating job transfer logic, their behaviors can be quietly embedded in the original process chain without triggering explicit errors, forming a deceptive "legal behavior chain" disguise;

[0004] This type of interference based on behavior injection can not only directly affect the calculation basis of key personnel decisions, but also, due to the lack of a unified behavior path redundancy verification structure, make it difficult for existing data consistency verification and error detection mechanisms to trace and identify or effectively intercept between heterogeneous systems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a human resource data security management method and system. By constructing a behavior trajectory tensor and a path redundancy projection space for the human resource business process, a cross-system personnel operation chain structure deviation detection mechanism is implemented to solve the problem that it is difficult to identify and verify the forged behavior chain during the consistency window period between heterogeneous human resource systems proposed in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solution: A human resource data security management method, comprising:

[0007] Obtain behavior log data, and after parsing the behavior dimension of the behavior log data, obtain a set of behavior event quadruples including behavior type, executor identity, execution context, and timestamp fields;

[0008] After performing a combined recoding operation of system module marking, execution node marking, and context nesting level marking on the set of behavior event quadruples, obtain an initial behavior vector sequence;

[0009] By performing time-series window sliding and behavior type clustering operations on the initial behavior vector sequence, continuous sub-behavior segments are obtained as the input source for multi-dimensional behavior sub-tensors; multi-axis alignment operations in the organizational structure dimension, system source dimension, and responsibility domain dimension are performed on the multi-dimensional behavior sub-tensors to obtain a set of behavior tensors with consistent structural levels;

[0010] Multi-process coupling operations and semantic label annotation operations are performed on the set of behavior tensors with consistent structural levels to obtain a unified behavior trajectory tensor body with process identifiers and context semantic annotations;

[0011] By constructing a path basis set and a behavior specification matrix, redundant projection and deviation analysis are performed on the behavior trajectory tensor body to generate a projection residual and deviation intensity map for anomaly recognition.

[0012] In a preferred embodiment, using the process meta-information and institutional rule parameters associated with the behavior trajectory tensor body, path parsing operations are performed to obtain a set of legal personnel behavior process paths, and each path in the set of legal personnel behavior process paths consists of step sequence, role constraints, and boundary states;

[0013] Path vectorization operations and standard state interpolation operations are performed on the set of legal personnel behavior process paths to obtain a path basis set, and each path basis in the path basis set is a state transition vector; dimension standardization and sequential dependency encoding operations are performed on the path basis set to obtain a behavior specification matrix, and each column in the behavior specification matrix represents a structured encoding expression of an accepted behavior path; the unified behavior trajectory tensor body is input into the behavior specification matrix, and multi-dimensional linear projection operations are performed to obtain a projection residual tensor, which is used to reflect the deviation intensity between the behavior trajectory and each path basis;

[0014] Dimension aggregation and anomaly degree normalization operations are performed on the projection residual tensor to obtain a deviation intensity map, and each map node is used to reflect the anomaly weight in the path dimension.

[0015] In a preferred embodiment, spatial gradient scanning and cross-dimensional linkage mode extraction operations are performed on the deviation intensity map to obtain a preliminary clustering result of the perturbation trajectory, and the output is a cluster of perturbation trajectories grouped by perturbation position, perturbation density, and behavior feature encoding;

[0016] Path correspondence mapping and anomaly source alignment operations are performed on the cluster of perturbation trajectories to obtain path perturbation classification identifiers, and the path perturbation classification identifiers divide the perturbation patterns into structure injection type, local perturbation type, and path mixed deformation type; the cluster of perturbation trajectories classified as the structure injection type is input into the jump node extraction and execution order consistency analysis module, and a set of suspected forged nodes and their associated behavior segment ranges are output.

[0017] In a preferred embodiment, the cluster of perturbation trajectories classified as the local perturbation class is input into the semantic consistency comparison and context causal conflict determination process, and the output is a perturbation node group, which is used to identify potential behavior errors or policy adjustment interventions; perform clustering center projection and adjacent legal path matching operations on all classified perturbation trajectory clusters to obtain a candidate correction path set and its supporting correction suggestion vector for subsequent in-tolerance trajectory correction.

[0018] In a preferred embodiment, the candidate correction path set and the original trajectory vector sequence are input into the deviation comparison process, and the path element difference analysis operation is performed, and the output is a path correction residual vector, which is used to record the node positions and behavior categories that need to be added, deleted, or replaced;

[0019] Perform context retention strategy calculation and behavior consistency rule filtering operations on the path correction residual vector, and the output is a behavior trajectory reconciliation plan, which includes execution order adjustment and role replacement configuration;

[0020] Perform legal path coverage verification and residual threshold evaluation operations on the behavior trajectory reconciliation plan, and the output is the final corrected trajectory. If the trajectory error is less than the preset policy threshold, it is determined that the repair is successful.

[0021] In a preferred embodiment, the successfully repaired final corrected trajectory enters the data archiving and snapshot mapping process, and the trajectory encapsulation and snapshot generation operations are performed, and the output is an adaptive trajectory snapshot, which is used for subsequent trajectory history comparison and behavior chain traceability audit;

[0022] Input the behavior trajectory with correction failure or residual exceeding the limit into the exception management process, perform the trajectory freezing and manual approval request binding operations, and the output is a pool of pending exception behaviors for subsequent manual intervention.

[0023] In a preferred embodiment, perform trajectory difference extraction operations on the trajectory snapshot data with correction completed or marked as abnormal, and the output is an evolution offset vector set, which is used to represent new variants of process behaviors; input the evolution offset vector set into the frequency analysis and cross-process impact clustering process, and the output is a candidate behavior evolution pattern, which is used to identify reasonable new trends in potential personnel processes;

[0024] Perform legality screening and residual regression verification operations on the candidate behavior evolution pattern, and the output is a set of variant paths, which is used as the basis for incremental expansion of the redundant path structure;

[0025] Merge the variant path set with the original path base set into the input path of the merging and singularity elimination process, output a new redundant path matrix, and form an updated behavior specification space; perform dynamic fault tolerance threshold re - evaluation and path confidence adjustment operations on the updated behavior specification space, and output a reconstructed redundant behavior model.

[0026] A human resource data security management system, including an event extraction module, a vector encoding module, a structure alignment module, a trajectory fusion module, and a redundancy mapping module;

[0027] The event extraction module is used to obtain behavior log data. After the behavior dimension is analyzed, a set of behavior event quadruples containing behavior type, executor identity, execution context, and timestamp fields is obtained;

[0028] The vector encoding module is used to perform a combined re - encoding operation of system module marking, execution node marking, and context nesting level marking on the set of behavior event quadruples to obtain an initial behavior vector sequence;

[0029] The structure alignment module obtains continuous sub - behavior segments by performing time - series window sliding and behavior type clustering operations on the initial behavior vector sequence, which are used as the input source of multi - dimensional behavior sub - tensors; perform multi - axis alignment operations on the multi - dimensional behavior sub - tensors in the organizational structure dimension, system source dimension, and responsibility domain dimension to obtain a set of behavior tensors with consistent structure levels;

[0030] The trajectory fusion module is used to perform multi - process coupling operations and semantic label annotation operations on the set of behavior tensors with consistent structure levels to obtain a unified behavior trajectory tensor body with process identifiers and context semantic annotations;

[0031] The redundancy mapping module constructs a path base set and a behavior specification matrix, performs redundant projection and deviation analysis on the behavior trajectory tensor body, and generates projection residuals and deviation intensity maps for anomaly identification.

[0032] The technical effects and advantages of the present invention:

[0033] 1. By constructing a behavior trajectory tensor and a path redundancy projection space, identify the forged behavior paths secretly injected during the cross - system consistency window period, and realize the structural - level verification and traceability of the disguised legal behavior chain;

[0034] 2. By introducing the orthogonal mapping mechanism of the path base set and the behavior specification matrix, map any behavior trajectory into a multi - path combination space, and realize the consistency anomaly detection of the personnel process in an asynchronous environment;

[0035] 3. By generating an abnormal heat map in the path dimension, it is possible to identify the concentrated distribution area of collective disturbance behaviors without explicit rule definition, thereby locating potential risk dimensions and improving the detection spatial resolution.

[0036] 4. Through the trajectory difference comparison and context consistency filtering strategy, it is possible to achieve tolerance repair at the behavior chain granularity level, effectively offset the interference of non-structural fluctuations on the trajectory judgment result, and ensure the logical continuity of the behavior chain.

[0037] 5. By introducing an evolutionary offset tensor and a dynamic redundant path update mechanism, the canonical space structure can be adaptively adjusted, which is beneficial to long-term adaptation to process evolution and attack pattern changes, and improves the self-healing ability and coverage breadth of the model. Description of the Drawings

[0038] Figure 1 It is a flowchart of the method steps of the present invention.

[0039] Figure 2 It is a schematic diagram of the system modules of the present invention.

[0040] Figure 3 It is a flowchart of the behavior trajectory construction stage of the present invention.

[0041] Figure 4 It is a flowchart of the path projection and residual generation stage of the present invention.

[0042] Figure 5 It is a flowchart of the deviation detection and clustering classification stage of the present invention.

[0043] Figure 6 It is a flowchart of the path correction and trajectory encapsulation stage of the present invention.

[0044] Figure 7 It is a flowchart of the structure evolution and model update stage of the present invention. Detailed Embodiments

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0046] Referring to the attached Figure 1-7 drawings, a human resource data security management method according to an embodiment of the present invention includes:

[0047] Obtain behavioral log data. After parsing the behavioral dimensions of the behavioral log data, a set of quadruple behavioral events containing fields such as behavior type, executor identity, execution context, and timestamp is obtained. Behavioral log data refers to operation records with audit significance triggered by users or system entities within a specific time in the human resources system, usually including content such as behavior type, approval, modification, submission, execution subject, target object, occurrence time, module where it is located, and context status, etc.

[0048] After performing a combined recoding operation of system module marking, execution node marking, and context nested level marking on the set of quadruple behavioral events, an initial behavioral vector sequence is obtained.

[0049] By performing time series window sliding and behavior type clustering operations on the initial behavioral vector sequence, continuous sub-behavior segments are obtained as the input source for multi-dimensional behavioral sub-tensors. Perform multi-axis alignment operations on the multi-dimensional behavioral sub-tensors in the organizational structure dimension, system source dimension, and responsibility domain dimension to obtain a set of behavior tensors with consistent structural levels.

[0050] Perform multi-process coupling operations and semantic label annotation operations on the set of behavior tensors with consistent structural levels to obtain a unified behavior trajectory tensor body with process identifiers and context semantic annotations.

[0051] By constructing a path basis set and a behavior specification matrix, perform redundant projection and deviation analysis on the behavior trajectory tensor body to generate a projection residual and deviation intensity map for anomaly recognition.

[0052] Utilize the process meta-information and institutional rule parameters associated with the behavior trajectory tensor body to perform path parsing operations to obtain a set of legal personnel behavior process paths. Each path in the set of legal personnel behavior process paths consists of step sequence, role constraints, and boundary states.

[0053] Perform path vectorization operations and standard state interpolation operations on the set of legal personnel behavior process paths to obtain a path basis set. Each path basis in the path basis set is a state transition vector. Perform dimension standardization and sequential dependency encoding operations on the path basis set to obtain a behavior specification matrix. Each column in the behavior specification matrix represents a structured coding expression of an accepted behavior path. Input the unified behavior trajectory tensor body into the behavior specification matrix and perform multi-dimensional linear projection operations to obtain a projection residual tensor, which is used to reflect the deviation intensity between the behavior trajectory and each path basis.

[0054] Perform dimension aggregation and anomaly degree normalization operations on the projection residual tensor to obtain a deviation intensity map. Each map node is used to reflect the anomaly weight in the path dimension.

[0055] Perform a spatial gradient scan and a cross-dimensional linkage mode extraction operation on the deviation intensity map to obtain a preliminary clustering result of the perturbation trajectories, and output perturbation trajectory clusters grouped by perturbation location, perturbation density, and behavior feature encoding;

[0056] Perform a path correspondence mapping and an abnormal source alignment operation on the perturbation trajectory clusters to obtain path perturbation classification identifiers, which divide the perturbation modes into structure injection type, local perturbation type, and path mixed deformation type; Input the perturbation trajectory clusters classified as the structure injection type into the jump node extraction and execution order consistency analysis module, and output a set of suspected forged nodes and the associated range of behavior segments.

[0057] Input the perturbation trajectory clusters classified as the local perturbation type into the semantic consistency comparison and context causal conflict determination process, and output a perturbation node group, which is used to identify potential behavior errors or policy adjustment interventions; Perform a clustering center projection and a neighboring legal path matching operation on all classified perturbation trajectory clusters to obtain a set of candidate correction paths and the accompanying correction suggestion vectors for subsequent in-tolerance trajectory correction.

[0058] Input the set of candidate correction paths and the original trajectory vector sequence into the deviation comparison process, perform a path element difference analysis operation, and output a path correction residual vector, which is used to record the node positions and behavior categories that need to be added, deleted, or replaced;

[0059] Perform a context preservation strategy calculation and a behavior consistency rule filtering operation on the path correction residual vector, and output a behavior trajectory reconciliation plan, which includes execution order adjustment and role replacement configuration;

[0060] Perform a legal path coverage verification and a residual threshold evaluation operation on the behavior trajectory reconciliation plan, and output the final corrected trajectory. If the trajectory error is less than the preset policy threshold, it is determined that the repair is successful.

[0061] Input the finally corrected trajectory with successful repair into the data archiving and snapshot mapping process, perform trajectory encapsulation and snapshot generation operations, and output an adaptive trajectory snapshot, which is used for subsequent trajectory history comparison and behavior chain traceability audit;

[0062] Input the behavior trajectories with failed correction or over-limit residuals into the abnormal management process, perform a trajectory freezing and artificial approval request binding operation, and output a pool of pending abnormal behaviors for subsequent manual intervention.

[0063] Perform a trajectory difference extraction operation on the trajectory snapshot data that has been corrected or marked as abnormal, and output an evolution offset vector set, which is used to represent the new variants of the process behavior; input the evolution offset vector set into the frequency analysis and cross-process impact clustering process, and output candidate behavior evolution patterns, which are used to identify reasonable new trends in potential personnel processes;

[0064] Perform a legality screening and residual regression verification operation on the candidate behavior evolution patterns, and output a set of variant paths, which is used as the basis for the incremental expansion of the redundant path structure;

[0065] Input the set of variant paths and the original path basis set into the path merging and singularity elimination process, and output a new redundant path matrix to form an updated behavior specification space; perform a dynamic fault tolerance threshold re-estimation and path confidence adjustment operation on the updated behavior specification space, and output a reconstructed redundant behavior model, which is used to improve the detection sensitivity and self-healing ability of the overall model;

[0066] It should be further explained as a whole that by mapping the unified behavior trajectory tensor body to the behavior specification space, the structural matching degree between the trajectory data and the legal path can be identified; in practical applications, first construct a path basis matrix composed of legal path encodings; then perform an orthogonal projection calculation on the behavior trajectory to obtain the structural fitting result of the trajectory in the specification space; finally, calculate the residual tensor between the original trajectory and the projection result, which is used as the input for subsequent anomaly identification and correction;

[0067] Define as the path basis set, and the path basis set represents a set of legal paths of personnel processes encoded by standardization;

[0068] where b (k) represents the vector encoding result of the kth path (unit: encoding factor / dimension); d is the encoding factor dimension, and the encoding factor dimension represents the representation dimension of each path (unit: encoding dimension); p is the number of paths (unit: piece); represents that this vector belongs to the d-dimensional real space;

[0069] For the construction of the behavior specification matrix, assume M is the behavior specification matrix, and the behavior specification matrix is formed by concatenating p legal path bases by columns; where represents a real matrix with d rows and p columns; M ij represents the value of the ith encoding dimension in the jth path in the behavior specification matrix (unit: encoding factor); in addition, the column space Col(M) of M defines the entire redundant projection space;

[0070] The behavioral trajectory tensor is defined as: where T is the unified behavioral trajectory tensor body; t (i) is the vector representation of the i-th trajectory in the encoding dimension (unit: encoding factor), forming a row; N is the total number of trajectory records (unit: number of records); is the real number space;

[0071] For the orthogonal projection operation of the trajectory, assume p (i) is the projection vector of the i-th trajectory in the path canonical space (unit: encoding factor);

[0072] where is the transpose of the behavior specification matrix, with dimension p×d; is the inverse of the Gram matrix formed by legal paths, used for the orthogonal calibration of the standard path group (unit: reciprocal of encoding factor); This projection operation is used to embed any trajectory vector into the combined space of all legal paths to obtain the most approximate matching trajectory form;

[0073] Immediately followed by the calculation of the projection residual,

[0074] where r (i) is the projection residual vector of the i-th trajectory, and the projection residual vector represents the deviation of its true behavioral trajectory from the combination of all legal paths (unit: encoding factor); ·t (i) is the original trajectory; p (i) is the trajectory fitting result after embedding into the path space; Among them, the magnitude of the residual reflects the degree of structural abnormality, which is the basic metric for anomaly detection;

[0075] Construct a non-linear measure of the structural offset, define δ (i) as the non-linear structural offset of the i-th trajectory (unit: encoding factor);

[0076]

[0077] where is the deviation value of the j-th dimension of the residual vector (unit: encoding factor); ln(1+exp(·)) is the Softplus function, and the Softplus function is used to non-linearly amplify the deviation value so that small deviations will not be ignored; This offset is used as a robust indicator of the behavioral chain structure offset to avoid misjudging short-term fluctuations as anomalies; represents the set of all real numbers greater than or equal to 0, that is, non-negative real numbers;

[0078] Finally, the anomaly heat map of the path dimension; [0,1]; where θ jis the abnormal heat index on the j-th path dimension (unit: coding factor); γ is the behavior perturbation reinforcement index (unit: dimensionless, used for non-linearly amplifying the weight); tanh(·) is the hyperbolic tangent function, and the hyperbolic tangent function in the formula is used to normalize the abnormal heat to the interval [0, 1]; where θ j The larger it is, it indicates that the j-th dimensional path coding shows a concentrated deviation in most trajectories, and it is the key dimension where abnormal structural features are concentrated; all θ j constitute the heat vector used to generate the path deviation atlas; N is the total number of behavioral trajectories participating in the calculation;

[0079] Based on the calculated projection residual vector of the trajectory, r (i) = t (i) - p (i) (indicating the difference between the trajectory and the projection path), further generate the structural deviation intensity index δ of each trajectory (i) , used to evaluate whether the trajectory substantially deviates from the standard behavioral path as a whole, and δ (i) will be used as the basic input for anomaly recognition and subsequent correction;

[0080] After the system runs for a period of time, its behavior specification matrix M may deviate from the historical state due to process evolution, structural adjustment or attack mutation. By comparing the historical and current path structures, construct the evolution tensor of the path basis, and then generate a new behavior specification matrix in combination with the non-linear update function to achieve the dynamic repair and enhancement of the redundant path space;

[0081] Define the path structure difference tensor, and formulate Δ ij as the structural change amplitude of the j-th path of the path basis in the i-th coding dimension (unit: coding factor);

[0082] where represents the path coding value of the i, j item in the current behavior specification matrix (unit: coding factor); is the path specification matrix of the previous time period (unit: coding factor); |·| represents taking the absolute value, focusing on the amplitude change and not distinguishing the direction;

[0083] The structural deviation amplification mapping is expressed as: E ij = ln(1 + exp(Δ ij )); E ij is the element of the structural deviation tensor, and the element of the structural deviation tensor is used to enhance the important fluctuations in the path change (unit: coding factor); the above formula suppresses small changes through the Softplus function and amplifies the expression ability of the structural perturbation;

[0084] Construct the path update weight generation function, and define σ ijTo update the weight factor (unit: dimensionless), which is used to reflect the degree of adoption of this structural change in path update; where σ ij ∈(0.5, 1), approaching 1 when the change is significant and the update is significant; approaching 0.5 for minor changes without affecting the structural stability;

[0085] Finally, the path specification space is adaptively updated. where represents the i, j unit in the updated behavior specification matrix (unit: coding factor); the above formula represents the current structure plus the weighted difference, indicating the response to path evolution; if the change is significant (σ≈1), the system quickly updates the path behavior template; if the change is not significant (σ≈0.5), the path structure is kept stable.

[0086] A human resource data security management system includes an event extraction module, a vector coding module, a structure alignment module, a trajectory fusion module, and a redundancy mapping module;

[0087] The event extraction module is used to obtain behavior log data, and after parsing the behavior dimensions, a set of behavior event quadruples containing behavior type, executor identity, execution context, and timestamp fields is obtained;

[0088] The vector coding module is used to perform a combined recoding operation of system module marking, execution node marking, and context nesting level marking on the set of behavior event quadruples to obtain an initial behavior vector sequence;

[0089] The structure alignment module obtains continuous sub - behavior segments as the input source of multi - dimensional behavior sub - tensors by performing time - series window sliding and behavior type clustering operations on the initial behavior vector sequence; and performs multi - axis alignment operations in the organizational structure dimension, system source dimension, and responsibility domain dimension on the multi - dimensional behavior sub - tensors to obtain a set of behavior tensors with consistent structural levels;

[0090] The trajectory fusion module is used to perform multi - process coupling operations and semantic label annotation operations on the set of behavior tensors with consistent structural levels to obtain a unified behavior trajectory tensor body with process identifiers and context semantic annotations;

[0091] The redundancy mapping module generates a projection residual and deviation intensity map for anomaly recognition by constructing a path basis set and a behavior specification matrix and performing redundancy projection and deviation analysis on the behavior trajectory tensor body.

[0092] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A human resource data security management method, comprising: Obtaining behavior log data, and after parsing the behavior dimension of the behavior log data, obtaining a set of behavior event quadruples including behavior type, executor identity, execution context, and timestamp fields; After performing a combined recoding operation of system module marking, execution node marking, and context nesting level marking on the set of behavior event quadruples, obtaining an initial behavior vector sequence; It is characterized in that: By performing time series window sliding and behavior type clustering operations on the initial behavior vector sequence, obtaining continuous sub-behavior segments as the input source of the multi-dimensional behavior sub-tensor; performing multi-axis alignment operations on the multi-dimensional behavior sub-tensor in the organizational structure dimension, system source dimension, and responsibility domain dimension, obtaining a set of behavior tensors with consistent structural levels; Performing a multi-process coupling operation and a semantic label annotation operation on the set of behavior tensors with consistent structural levels, obtaining a unified behavior trajectory tensor body with process identifiers and context semantic annotations; By constructing a path basis set and a behavior specification matrix, performing redundant projection and deviation analysis on the behavior trajectory tensor body, generating a projection residual and a deviation intensity map for anomaly recognition.

2. A human resource data security management method according to claim 1, characterized in that: Utilizing the process meta-information and institutional rule parameters associated with the behavior trajectory tensor body, performing a path parsing operation, obtaining a set of legal personnel behavior process paths, and each path in the set of legal personnel behavior process paths consists of step sequence, role constraint, and boundary state; Performing a path vectorization operation and a standard state interpolation operation on the set of legal personnel behavior process paths, obtaining a path basis set, and each path basis in the path basis set is a state transition vector; Performing a dimension standardization and sequential dependence encoding operation on the path basis set, obtaining a behavior specification matrix, and each column in the behavior specification matrix represents a structured encoding expression of an accepted behavior path; inputting the unified behavior trajectory tensor body into the behavior specification matrix, performing a multi-dimensional linear projection operation, obtaining a projection residual tensor, and the projection residual tensor is used to reflect the deviation intensity between the behavior trajectory and each path basis; Performing a dimension aggregation and anomaly degree normalization operation on the projection residual tensor, obtaining a deviation intensity map, and each map node is used to reflect the anomaly weight on the path dimension.

3. A human resource data security management method according to claim 2, characterized in that: Performing a spatial gradient scanning and cross-dimensional linkage mode extraction operation on the deviation intensity map, obtaining a preliminary clustering result of the perturbation trajectory, and outputting a perturbation trajectory cluster grouped by perturbation position, perturbation density, and behavior feature encoding; Performing a path correspondence mapping and anomaly source alignment operation on the perturbation trajectory cluster, obtaining a path perturbation classification identifier, and the path perturbation classification identifier divides the perturbation mode into a structure injection class, a local perturbation class, and a path mixed deformation class; inputting the perturbation trajectory cluster classified as the structure injection class into a jump node extraction and execution order consistency analysis module, and outputting a set of suspected forged nodes and the associated behavior segment range.

4. A human resource data security management method according to claim 3, characterized in that: Input the cluster of perturbation trajectories classified as the local perturbation class into the semantic consistency comparison and context causal conflict determination process, and output a group of perturbation nodes, which is used to identify potential behavior errors or policy adjustment interventions; perform clustering center projection and adjacent legal path matching operations on all classified clusters of perturbation trajectories to obtain a set of candidate correction paths and their supporting correction suggestion vectors for subsequent trajectory correction within tolerance.

5. A human resource data security management method according to claim 4, characterized in that: Input the set of candidate correction paths and the original trajectory vector sequence into the deviation comparison process, perform path element difference analysis operations, and output a path correction residual vector, which is used to record the node positions and behavior categories that need to be added, deleted, or replaced; Perform context retention strategy calculation and behavior consistency rule filtering operations on the path correction residual vector, and output a behavior trajectory reconciliation plan, which includes execution order adjustment and role replacement configuration; Perform legal path coverage verification and residual threshold evaluation operations on the behavior trajectory reconciliation plan, and output the final corrected trajectory. If the trajectory error is less than the preset policy threshold, it is determined that the repair is successful.

6. A human resource data security management method according to claim 5, characterized in that: Input the finally corrected trajectory with successful repair into the data sealing and snapshot mapping process, perform trajectory encapsulation and snapshot generation operations, and output an adaptive trajectory snapshot, which is used for subsequent trajectory history comparison and behavior chain traceability audit; Input the behavior trajectory with failed correction or residual exceeding the limit into the exception management process, perform trajectory freezing and binding of manual approval requests, and output a pool of abnormal behaviors to be audited for subsequent manual intervention.

7. A human resource data security management method according to claim 6, characterized in that: Perform trajectory difference extraction operations on the trajectory snapshot data with correction completed or marked as abnormal, and output a set of evolution offset vectors, which is used to represent new variants of process behaviors; input the set of evolution offset vectors into the frequency analysis and cross-process impact clustering process, and output candidate behavior evolution patterns, which are used to identify reasonable new trends of potential personnel processes; Perform legality screening and residual regression verification operations on the candidate behavior evolution patterns, and output a set of variant paths, which is used as the basis for incremental expansion of the redundant path structure; Input the set of variant paths and the original path basis set into the path merging and singularity elimination process, and output a new redundant path matrix to form an updated behavior specification space; Perform dynamic fault tolerance threshold re-estimation and path confidence adjustment operations on the updated behavior specification space, and output a reconstructed redundant behavior model.

8. A human resource data security management system, including an event extraction module, a vector encoding module, a structure alignment module, a trajectory fusion module, and a redundant mapping module, characterized in that: The event extraction module is used to obtain behavior log data, and after parsing the behavior dimension, a set of behavior event quadruples including behavior type, executor identity, execution context, and timestamp fields is obtained; The vector encoding module is used to perform a combined recoding operation of system module marking, execution node marking, and context nesting level marking on the set of behavioral event quadruples, and then obtain an initial behavioral vector sequence; The structure alignment module obtains continuous sub-behavior segments by performing time-series window sliding and behavior type clustering operations on the initial behavioral vector sequence, which serve as the input source for multi-dimensional behavioral sub-tensors; Perform multi-axis alignment operations on the multi-dimensional behavioral sub-tensors in the organizational structure dimension, system source dimension, and responsibility domain dimension to obtain a set of behavioral tensors with consistent structural levels; The trajectory fusion module is used to perform multi-process coupling operations and semantic label annotation operations on the set of behavioral tensors with consistent structural levels to obtain a unified behavioral trajectory tensor body with process identifiers and context semantic annotations; The redundancy mapping module constructs a path basis set and a behavior specification matrix, performs redundancy projection and deviation analysis on the behavioral trajectory tensor body, and generates a projection residual and deviation intensity map for anomaly recognition.

Images