An enterprise accounting management system based on enterprise data security audit

Abstract

The present invention discloses an enterprise accounting management system based on enterprise data security audit, and relates to the technical field of accounting management systems. The enterprise accounting management system based on enterprise data security audit includes: a file upload module, a manual permission setting module, an automatic identification and permission division module, a permission change audit module, a permission display and confirmation module, a permission management and access control module, and a log and audit module; a comprehensive and efficient enterprise accounting data security management system is constructed through multiple modules, which can automatically identify file types and intelligently divide permission levels, while supporting manual permission setting to meet specific needs, thereby improving the flexibility and accuracy of permission management. In addition, the system has also established a strict permission change review process and a regular review mechanism to ensure that the file permission settings always match the importance and sensitivity of the file, effectively reducing the risk of data leakage.

Description

Technical Field

[0001] The present invention relates to the technical field of accounting management systems, and in particular to an enterprise accounting management system based on enterprise data security audit. Background Art

[0002] The financial management system of an enterprise is a system that comprehensively manages all aspects of financial management, such as analysis, forecasting, planning, control, and supervision, through information management and network communication systems. It is a value-oriented information system based on accounting management activities, usually running in a computer hardware and network environment, and using modern information processing technology to collect, store, process, and transmit accounting information to complete tasks such as accounting, supervision, management, and decision-making assistance.

[0003] For example, a financial management system and method disclosed in publication number CN115587899A includes administrator login, cashier / accountant login, financial director login and general manager login. The administrator login includes a management module and an authority module. The cashier / accountant login includes initial settings, material entry, primary management and transmission modules. Primary management includes voucher management and account book management. The financial director login includes a receiving module, senior management and financial analysis. Senior management includes material summary. The material summary includes transaction summary, income summary and cost summary. The general manager management includes a query module.

[0004] As shown in the above technology, the current financial management system of an enterprise, because it involves some relatively important data, will design a permission module to set permissions in a hierarchical manner, open different permissions for employees of different levels to view file data of different permissions, and such files need to set permissions when uploading. Currently, the permissions are determined by the uploader when uploading, or the administrator divides the permissions. The permission module of the current financial management system can only set user permissions, but cannot automatically divide file permissions. For large enterprises with large data volumes, the workload of relying on employees to handle it by themselves is large and prone to errors. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an enterprise accounting management system based on enterprise data security audit, which solves the problem that the permission module of the current financial management system can only set user permissions but cannot automatically divide file permissions. For large enterprises with large amounts of data, manual processing is labor-intensive and prone to errors.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an enterprise accounting management system based on enterprise data security audit, comprising:

[0007] The file upload module receives new files uploaded by the accountant or other users and uploads new files U to the accountant, and records the uploader information P user And file basic information B U (such as file name, file type, upload time, etc.), and set the variable P manual False, indicating that permissions are not manually set in the initial state;

[0008] Manual permission setting module allows uploaders to manually set file permissions and update P manual True, and record the manually set permission level L manual If the permissions are not set manually, P manual is False;

[0009] Automatic identification and permission division module, used to pre-process uploaded files, using file type identification model M type Identify the file type U , and use the permission level classification model M level Divide permissions into different levels;

[0010] The permission change review module records permission change information, establishes approval processes, and regularly reviews the permission levels of files;

[0011] The permission display and confirmation module is used to display the adjusted permission level to the uploader or administrator with higher permissions for confirmation and adjustment;

[0012] The permission management and access control module sets access permissions for files according to the final permission level and manages the user's access permissions;

[0013] The log and audit module records system operation information including file uploads, permission changes, and access logs, and provides log query and audit functions for subsequent reference and auditing.

[0014] Preferably, the file type identification model M type And the authority level division model M level The build steps include:

[0015] A1. Data preparation and preprocessing:

[0016] A11. Assume that set D represents the complete set of enterprise file data, where each element d i Represents a file data;

[0017] A12. For each element d in set D i Perform preprocessing, including denoising, image enhancement, text extraction, etc., to obtain the preprocessed file data set D';

[0018] A2. Feature extraction and selection:

[0019] A21. Extract features from the preprocessed file data set D' and set the feature vector F i Indicates file d i The feature set of

[0020] A22. Use feature selection methods (such as chi-square test, mutual information, etc.) to filter out features that are useful for file type identification and permission level classification, and obtain the filtered feature vector F' i ;

[0021] A3. Model training and optimization:

[0022] A31. Set the data set T as the filtered feature vector set and its corresponding file type and permission level label, and the data set T includes a training data set and a test data set;

[0023] A32. Use machine learning algorithms (such as support vector machines, decision trees, etc.) to train the training data set T to obtain a file type recognition model M. type and permission level division model M level ;

[0024] A33. Set the accuracy A as the performance of the model on the test data set, and optimize the model through methods such as cross-validation until A reaches a satisfactory level.

[0025] Preferably, step A31 specifically includes: randomly dividing a training data set T from the filtered feature vectors train And the test data set T test , the ratio is usually 7:3 or 8:2; set the training data set T train is {(F1′, Y1), (F2′, Y2), …, (F m ′,Y m )}, where F i ′ represents the filtered feature vector, Yi represents the corresponding file type and permission level label (can be multiple labels);

[0026] Step A32 specifically includes: using the training data set T train File type recognition model M type Perform training to obtain the trained model M type ′;

[0027] Using the training dataset T train Model M for classifying authority levels level Perform training to obtain the trained model M level ′;

[0028] Step A33 specifically includes: setting the accuracy A as the model’s accuracy in the test data set T test Based on the performance on the training set, use cross-validation (such as k-fold cross-validation) to evaluate the model to reduce the risk of overfitting and find the optimal model parameters; according to the results of cross-validation, adjust the model parameters and retrain the model until the accuracy A reaches a satisfactory level.

[0029] Preferably, finding the optimal model parameters means selecting the model with the highest accuracy and the strongest generalization ability among multiple candidate models as the final file type recognition model. and permission level division model Among them, accuracy refers to the proportion of correct predictions made by the model on the test set; generalization ability refers to the model's ability to predict new data (unseen data), and the generalization ability of the model is indirectly evaluated by the average accuracy and standard deviation of cross-validation;

[0030] The final model is trained using the entire training dataset (or a repartitioned larger training dataset) to improve the stability and accuracy of the model.

[0031] Preferably, the accuracy rate A is calculated as follows:

[0032]

[0033] Among them, δ(M(F i ′), Y i ′) indicates that when M(F i ′)=Y i ′ takes 1, otherwise takes 0; M(F i ′) represents the model for the feature vector F i ′’s prediction result; i ′ represents the eigenvector F i ′ is the true label; n represents the number of samples in the test data set;

[0034] Cross-validation step: The training data set T train Divide into k non-overlapping subsets; for each subset, use the remaining k-1 subsets for training, and use the current subset for testing; calculate the average accuracy of k tests as the cross-validation accuracy.

[0035] Preferably, in the automatic identification and authority division module, the uploaded file U is pre-processed and output as U'. manual If False, the permission level classification model M is used. level Divide U' into permission levels and obtain the preliminary permission level L initial ;

[0036] Set the rule set R to include permission division rules based on file type, keyword, uploader identity and file content sensitivity;

[0037] According to the rule set R, the permission level L is initially divided initial Make adjustments and get the adjusted permission level L final .

[0038] Preferably, the work content of the permission change review module specifically includes:

[0039] Authorization change record: When the authority level is changed from L initial Adjust to L final During the process, the system records the detailed information of each permission change, including the time of change, the person who changed, and the permission level before and after the change, and stores the detailed information in a secure log system for subsequent audit and tracing;

[0040] Change approval process: For non-manual permission changes, i.e. changes made automatically by the system or by administrators, a strict approval process is established; change requests must be reviewed and approved by the appropriate authority before they can take effect;

[0041] Regular permission review: To ensure that the file permission settings match the importance and sensitivity, the system needs to regularly review the file permission levels; if it is found during the review process that the permission settings do not comply with the rules, timely adjustments will be made; the review results will be recorded for subsequent reference and auditing.

[0042] Preferably, the work contents of the permission display and confirmation module specifically include:

[0043] The adjusted permission level L final The file is displayed to the uploader or an administrator with higher authority for confirmation and adjustment. During the confirmation period, the file is temporarily stored and inaccessible. If no confirmation or adjustment operation is performed within the set time, it is confirmed by default, and the file is exited from the temporary storage state for storage, at which time it can be accessed.

[0044] If P manual is True, then L final Equal to L manual .

[0045] Preferably, in the permission display and confirmation module, if the reviewer still insists on modifying the permissions back to the original permissions after the system has divided the permissions, the variable P is set revert Indicates whether the manual modification is to restore the original permissions. The initial value is False. The reason for modification R is evaluated using a learning algorithm (such as reinforcement learning, gradient boosting tree, etc.). reason To study;

[0046] Assume M learnRepresents the learning algorithm, accepts the modification reason R reason As input, and output the updated permission division rule set R′;

[0047] R′=M learn (R,R reason );

[0048] Among them, R is the original set of rules for authority division, R reason is the reason for modification provided manually, M learn is the learning algorithm, R′ is the updated rule set for authority division;

[0049] According to the updated permission partition rule set R′ or R multi ′, automatically identify and divide permissions for subsequent uploaded files.

[0050] Preferably, the work contents of the authority management and access control module specifically include:

[0051] According to the final permission level L final Set corresponding access permissions for file U;

[0052] Use the role-based access control model (RBAC) to manage user access rights and ensure that only users with corresponding permissions can access file U;

[0053] The corresponding relationship between permission level and access right is expressed as: final →Access permission set;

[0054] Among them, L final is the final permission level, and the access permission set is a list of user access permissions corresponding to this level.

[0055] The present invention provides an enterprise accounting management system based on enterprise data security audit. Compared with the prior art, it has the following beneficial effects:

[0056] 1. This enterprise accounting management system based on enterprise data security audit has built a comprehensive and efficient enterprise accounting data security management system through multiple modules. It can automatically identify file types and intelligently divide permission levels. It also supports manual setting of permissions to meet specific needs, improving the flexibility and accuracy of permission management. In addition, the system has established a strict permission change review process and a regular review mechanism to ensure that file permission settings always match the importance and sensitivity of the file, effectively reducing the risk of data leakage. Through the log and audit modules, the system can comprehensively record and track file operation history, providing strong support for subsequent auditing and tracing. Overall, the system has significantly improved the security and management efficiency of corporate accounting data.

[0057] 2. The enterprise accounting management system based on enterprise data security audit builds a file type recognition model M type and permission level division model M leve l. The system can automatically complete the steps of preprocessing, feature extraction, model training and optimization of enterprise file data. The two models can accurately identify the type of file and intelligently divide the permission level according to factors such as the content of the file and the identity of the uploader, thereby greatly improving the efficiency and accuracy of file management. In addition, by adopting evaluation methods such as cross-validation, the system can find the optimal model parameters to ensure that the model has good generalization ability and stability. This automatic identification and classification mechanism not only reduces manual intervention, but also reduces the risk of human error, providing strong technical support for enterprise data security audits. Overall, the implementation of this system has significantly improved the level of intelligence in enterprise file management.

[0058] 3. The enterprise accounting management system based on enterprise data security audit not only ensures the transparency and traceability of permission adjustment by introducing the permission display and confirmation module, but also effectively prevents unconfirmed files from being accessed by setting a temporary storage status, thereby improving data security. At the same time, when the auditor wants to modify the permission, the system can record and learn the reason for the modification, and use the learning algorithm to continuously optimize the permission division rules, so that the subsequent file permission division is more accurate and reasonable. This mechanism not only respects manual judgment, but also continuously improves itself through machine learning, realizes the intelligent upgrade of permission management, and provides a strong guarantee for the security and efficient use of enterprise data. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a system principle block diagram of the present invention;

[0060] Figure 2 It is a schematic diagram of the overall process of the present invention;

[0061] Figure 3 The figure is a schematic diagram of the model construction process of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] The present invention provides the following three technical solutions:

[0064] Figure 1-Figure 2The first embodiment is shown: an enterprise accounting management system based on enterprise data security audit, comprising:

[0065] The file upload module receives new files uploaded by the accountant or other users and uploads new files U to the accountant, and records the uploader information P user And file basic information B U (such as file name, file type, upload time, etc.), and set the variable P manual False, indicating that permissions are not manually set in the initial state;

[0066] Manual permission setting module allows uploaders to manually set file permissions and update P manual True, and record the manually set permission level L manual If the permissions are not set manually, P manual is False;

[0067] Automatic identification and permission division module, used to pre-process uploaded files, using file type identification model M type Identify the file type U , and use the permission level classification model M level Divide the permission level; the uploaded file U is preprocessed and output as U'. manual If False, the permission level classification model M is used. level Divide U' into permission levels and obtain the preliminary permission level L initial ;

[0068] Set the rule set R to include permission division rules based on file type, keyword, uploader identity and file content sensitivity. initial Make adjustments and get the adjusted permission level L final ;

[0069] The permission change review module records permission change information, establishes approval processes, and regularly reviews the permission levels of files; specifically includes:

[0070] Authorization change record: When the authority level is changed from L initial Adjust to L final During the process, the system records the detailed information of each permission change, including the time of change, the person who changed, and the permission level before and after the change, and stores the detailed information in a secure log system for subsequent audit and tracing;

[0071] Change approval process: For non-manual permission changes, i.e. changes made automatically by the system or by administrators, a strict approval process is established; change requests must be reviewed and approved by the appropriate authority before they can take effect;

[0072] Regular permission review: To ensure that the permission settings of files match their importance and sensitivity, the system needs to regularly review the permission levels of files. If permission settings are found to be inconsistent with the rules during the review process, they should be adjusted in a timely manner. The review results should be recorded for subsequent reference and auditing.

[0073] The permission display and confirmation module is used to display the adjusted permission level to the uploader or administrator with higher permissions for confirmation and adjustment;

[0074] The permission management and access control module sets access permissions for files according to the final permission level and manages user access permissions, including:

[0075] According to the final permission level L final Set corresponding access permissions for file U;

[0076] Use the role-based access control model (RBAC) to manage user access rights and ensure that only users with corresponding permissions can access file U;

[0077] The corresponding relationship between permission level and access right is expressed as: final →Access permission set;

[0078] Among them, L final is the final permission level, and the access permission set is the user access permission list corresponding to this level;

[0079] The log and audit module records system operation information including file uploads, permission changes, and access logs, and provides log query and audit functions for subsequent reference and auditing.

[0080] This system has built a comprehensive and efficient enterprise accounting data security management system through multiple modules. It can automatically identify file types and intelligently divide permission levels. It also supports manual setting of permissions to meet specific needs, improving the flexibility and accuracy of permission management. In addition, the system has established a strict permission change review process and a regular review mechanism to ensure that file permission settings always match the importance and sensitivity of the file, effectively reducing the risk of data leakage. Through the log and audit modules, the system can comprehensively record and track file operation history, providing strong support for subsequent auditing and tracing. Overall, the system has significantly improved the security and management efficiency of corporate accounting data.

[0081] Figure 3 A second embodiment is shown, which is mainly different from the first embodiment in that: the file type recognition model M type And the authority level division model M level The build steps include:

[0082] A1. Data preparation and preprocessing:

[0083] A11. Assume that set D represents the complete set of enterprise file data, where each element d i Represents a file data;

[0084] A12. For each element d in set D i Perform preprocessing, including denoising, image enhancement, text extraction, etc., to obtain the preprocessed file data set D';

[0085] A2. Feature extraction and selection:

[0086] A21. Extract features from the preprocessed file data set D' and set the feature vector F i Indicates file d i The feature set of

[0087] A22. Use feature selection methods (such as chi-square test, mutual information, etc.) to filter out features that are useful for file type identification and permission level classification, and obtain the filtered feature vector F' i ;

[0088] A3. Model training and optimization:

[0089] A31. Set the data set T as the filtered feature vector set and its corresponding file type and permission level label. The data set T includes a training data set and a test data set. Specifically, randomly divide the training data set T from the filtered feature vector. train And the test data set T test , the ratio is usually 7:3 or 8:2; set the training data set T train is {(F1′, Y1), (F2′, Y2), …, (F m ′,Y m )}, where F i ′ represents the filtered feature vector, Yi represents the corresponding file type and permission level label (can be multiple labels);

[0090] A32. When using machine learning algorithms (such as support vector machines (SVMs) and decision trees (DTs), model parameters must be initialized first, such as the penalty parameter C and kernel function parameter γ of SVMs, or the depth d and splitting criteria of decision trees;

[0091] SVM parameters:

[0092] Penalty parameter C: controls the complexity and generalization ability of the model, and its value range is usually [0.1, 100];

[0093] Kernel function parameter γ: affects the performance of the model in high-dimensional space, and its value range is usually [0.001, 1];

[0094] Decision tree parameters:

[0095] Depth d: controls the growth depth of the tree to prevent overfitting. The value range depends on the complexity of the data set.

[0096] Split criteria: such as Gini coefficient, entropy, etc., used to select the best split feature;) Train the training data set T to obtain the file type recognition model M type And the authority level division model M level ; Use training data set T train File type recognition model M type Perform training to obtain the trained model M type ′;

[0097] Using the training dataset T train Model M for classifying authority levels level Perform training to obtain the trained model M level ′;

[0098] A33. Set the accuracy A as the model's accuracy in the test data set T test Based on the performance on the training set, use cross-validation (such as k-fold cross-validation) to evaluate the model to reduce the risk of overfitting and find the optimal model parameters; according to the results of cross-validation, adjust the model parameters and retrain the model until the accuracy A reaches a satisfactory level.

[0099] Finding the optimal model parameters means selecting the model with the highest accuracy and the strongest generalization ability among multiple candidate models as the final file type recognition model. and permission level division model Among them, accuracy refers to the proportion of correct predictions made by the model on the test set; generalization ability refers to the model's ability to predict new data (unseen data), and the generalization ability of the model is indirectly evaluated by the average accuracy and standard deviation of cross-validation;

[0100] The final model is trained using the entire training dataset (or a repartitioned larger training dataset) to improve the stability and accuracy of the model.

[0101] The calculation formula of the accuracy rate A is:

[0102]

[0103] Among them, δ(M(F i ′), Y i ′) indicates that when M(F i ′)=Y i′ takes 1, otherwise takes 0; M(F i ′) represents the model for the feature vector F i ′’s prediction result; i ′ represents the eigenvector F i ′ is the true label; n represents the number of samples in the test data set;

[0104] Cross-validation step: The training data set T train Divide into k non-overlapping subsets; for each subset, use the remaining k-1 subsets for training, and use the current subset for testing; calculate the average accuracy of k tests as the cross-validation accuracy.

[0105] Identify the model M by building the file type type and permission level division model M leve l. The system can automatically complete the steps of preprocessing, feature extraction, model training and optimization of enterprise file data. The two models can accurately identify the type of file and intelligently divide the permission level according to factors such as the content of the file and the identity of the uploader, thereby greatly improving the efficiency and accuracy of file management. In addition, by adopting evaluation methods such as cross-validation, the system can find the optimal model parameters to ensure that the model has good generalization ability and stability. This automatic identification and classification mechanism not only reduces manual intervention, but also reduces the risk of human error, providing strong technical support for enterprise data security audits. Overall, the implementation of this system has significantly improved the level of intelligence in enterprise file management.

[0106] The third implementation mode is mainly different from the first implementation mode in that the work content of the permission display and confirmation module specifically includes:

[0107] The adjusted permission level L final The file is displayed to the uploader or an administrator with higher authority for confirmation and adjustment. During the confirmation period, the file is temporarily stored and inaccessible. If no confirmation or adjustment operation is performed within the set time, it is confirmed by default, and the file is exited from the temporary storage state for storage, at which time it can be accessed.

[0108] If P manual is True, then L final Equal to L manual .

[0109] If the reviewer still insists on changing the permissions back to the original permissions after the system has divided the permissions, then set the variable P revert Indicates whether the manual modification is to restore the original permissions. The initial value is False. The reason for modification R is evaluated using a learning algorithm (such as reinforcement learning, gradient boosting tree, etc.). reason To study;

[0110] Assume Mlearn Represents the learning algorithm, accepts the modification reason R reason As input, and output the updated permission division rule set R′;

[0111] R′=M learn (R,R reason );

[0112] Among them, R is the original set of rules for authority division, R reason is the reason for modification provided manually, M learn is the learning algorithm, R′ is the updated rule set for authority division;

[0113] According to the updated permission partition rule set R′ or R multi ′, automatically identify and divide permissions for subsequent uploaded files.

[0114] By introducing the permission display and confirmation module, not only the transparency and traceability of permission adjustment are ensured, but also by setting the temporary storage status, unconfirmed files are effectively prevented from being accessed, thereby improving data security. At the same time, when the auditor wants to modify the permission, the system can record and learn the reason for the modification, and use the learning algorithm to continuously optimize the permission division rules, making the subsequent file permission division more accurate and reasonable. This mechanism not only respects human judgment, but also continuously improves itself through machine learning, realizing the intelligent upgrade of permission management, and providing a strong guarantee for the security and efficient use of enterprise data.

[0115] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0116] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0117] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An enterprise accounting management system based on enterprise data security audit, characterized in that: include: The file upload module receives new files uploaded by the accountant or other users and uploads new files U to the accountant, and records the uploader information P user And basic information of the file B U , and set the variable P manual False, indicating that permissions are not manually set in the initial state; Manual permission setting module allows uploaders to manually set file permissions and update P manual True, and record the manually set permission level L manual If the permissions are not set manually, P manual is False; Automatic identification and permission division module, used to pre-process uploaded files, using file type identification model M type Identify the file type U , and use the permission level classification model M level Divide permissions into different levels; The permission change review module records permission change information, establishes approval processes, and regularly reviews the permission levels of files; The permission display and confirmation module is used to display the adjusted permission level to the uploader or administrator with higher permissions for confirmation and adjustment; The permission management and access control module sets access permissions for files according to the final permission level and manages the user's access permissions; The log and audit module records system operation information including file uploads, permission changes, and access logs, and provides log query and audit functions for subsequent reference and auditing.

2. According to claim 1, the enterprise accounting management system based on enterprise data security audit is characterized by: The file type identification model M type and permission level division model M level The build steps include: A1. Data preparation and preprocessing: A11. Assume that set D represents the complete set of enterprise file data, where each element d i Represents a file data; A12. For each element d in set D i Perform preprocessing, including denoising, image enhancement, and text extraction, to obtain a preprocessed file data set D'; A2. Feature extraction and selection: A21. Extract features from the preprocessed file data set D' and set the feature vector F i Indicates file d i The feature set of A22. Use feature selection methods to filter out features that are useful for file type identification and permission level classification, and obtain the filtered feature vector F' i ; A3. Model training and optimization: A31. Set the data set T as the filtered feature vector set and its corresponding file type and permission level label, and the data set T includes a training data set and a test data set; A32. Use machine learning algorithms to train the training data set T to obtain a file type recognition model M type and permission level division model M level ; A33. Set the accuracy A as the performance of the model on the test data set, and optimize the model through cross-validation method until A reaches a satisfactory level.

3. According to claim 2, the enterprise accounting management system based on enterprise data security audit is characterized by: Step A31 specifically includes: randomly dividing the training data set T from the filtered feature vector train And the test data set T test ; Set the training data set T train is {(F1′, Y1), (F2′, Y2), …, (F m ′,Y m )}, where F i ′ represents the filtered feature vector, Yi represents the corresponding file type and permission level label; Step A32 specifically includes: using the training data set T train File type recognition model M type Perform training to obtain the trained model M type ′; Using the training dataset T train Model M for classifying authority levels level Perform training to obtain the trained model M level ′; Step A33 specifically includes: setting the accuracy A as the model’s accuracy in the test data set T test Based on the performance on the training set, cross-validation is used to evaluate the model to reduce the risk of overfitting and find the optimal model parameters. Based on the results of cross-validation, the model parameters are adjusted and the model is retrained until the accuracy A reaches a satisfactory level.

4. According to claim 3, the enterprise accounting management system based on enterprise data security audit is characterized by: Finding the optimal model parameters means selecting the model with the highest accuracy and the strongest generalization ability among multiple candidate models as the final file type recognition model. and permission level division model Among them, accuracy refers to the proportion of correct predictions made by the model on the test set; generalization ability refers to the model's ability to predict new data, and the generalization ability of the model is indirectly evaluated by the average accuracy and standard deviation of cross-validation; The final model is trained using the entire training dataset to improve the stability and accuracy of the model.

5. According to claim 3, the enterprise accounting management system based on enterprise data security audit is characterized by: The calculation formula of the accuracy rate A is: Among them, δ(M(F i ′), Y i ′) indicates that when M(F i ′)=Y i ′ takes 1, otherwise takes 0; M(F i ′) represents the model for the feature vector F i ′’s prediction result; i ′ represents the eigenvector F i ′ is the true label; n represents the number of samples in the test data set; Cross-validation step: The training data set T train Divide into k non-overlapping subsets; for each subset, use the remaining k-1 subsets for training, and use the current subset for testing; calculate the average accuracy of k tests as the cross-validation accuracy.

6. The enterprise accounting management system based on enterprise data security audit according to claim 1 is characterized by: In the automatic identification and authority division module, the uploaded file U is preprocessed and output as U'. manual If False, the permission level classification model M is used. level Divide U' into permission levels and obtain the preliminary permission level L initial ; Set the rule set R to include permission division rules based on file type, keyword, uploader identity and file content sensitivity; According to the rule set R, the permission level L is initially divided initial Make adjustments and obtain the adjusted permission level L final .

7. The enterprise accounting management system based on enterprise data security audit according to claim 6 is characterized by: The work content of the permission change review module specifically includes: Authorization change record: When the authority level is changed from L initial Adjust to L final During the process, the system records the detailed information of each permission change, including the time of change, the person who changed, and the permission level before and after the change, and stores the detailed information in a secure log system for subsequent audit and tracing; Change approval process: For non-manual permission changes, i.e. changes made automatically by the system or by administrators, a strict approval process is established; change requests must be reviewed and approved by the appropriate authority before they can take effect; Regular permission review: To ensure that the file permission settings match the importance and sensitivity, the system needs to regularly review the file permission levels; if it is found during the review process that the permission settings do not comply with the rules, timely adjustments will be made; the review results will be recorded for subsequent reference and auditing.

8. The enterprise accounting management system based on enterprise data security audit according to claim 7 is characterized by: The work content of the permission display and confirmation module specifically includes: The adjusted permission level L final The file is displayed to the uploader or an administrator with higher authority for confirmation and adjustment. During the confirmation period, the file is temporarily stored and inaccessible. If no confirmation or adjustment operation is performed within the set time, it is confirmed by default, and the file is exited from the temporary storage state for storage, at which time it can be accessed. If P manual is True, then L final Equal to L manual .

9. The enterprise accounting management system based on enterprise data security audit according to claim 8 is characterized by: In the permission display and confirmation module, if the reviewer still insists on changing the permissions back to the original permissions after the system has divided the permissions, the variable P is set. revert Indicates whether the manual modification is to restore the original permissions. The initial value is False. The learning algorithm is used to modify the reason R reason To study; Assume M learn Represents the learning algorithm, accepts the modification reason R reason As input, and output the updated permission division rule set R′; R′=M learn (R,R reason ); Among them, R is the original set of rules for authority division, R reason is the reason for modification provided manually, M learn is the learning algorithm, R′ is the updated rule set for authority division; According to the updated permission partition rule set R′ or R multi ′, automatically identify and divide permissions for subsequent uploaded files.

10. The enterprise accounting management system based on enterprise data security audit according to claim 6 is characterized by: The work content of the authority management and access control module specifically includes: According to the final permission level L final Set corresponding access permissions for file U; Use the role-based permission management model to manage user access rights, ensuring that only users with corresponding permissions can access the file U; The corresponding relationship between permission level and access right is expressed as: final →Access permission set; Among them, L final is the final permission level, and the access permission set is the same as L final The corresponding user access permission list.

Images