Enterprise tax payment sensitive data identification method, device, equipment and medium

Abstract

The invention discloses an enterprise tax sensitive data identification method and device, equipment and a medium, and relates to the technical field of data processing, and the method comprises the steps: obtaining tax data, carrying out the data division, and obtaining first tax data for model training and to-be-identified second tax data; performing data cleaning and processing on the first tax payment data by using a diffusion time estimation network, constructing a diffusion link, inputting the processed tax payment data into the diffusion link, outputting a vector, and performing posterior distribution calculation on the vector to obtain posterior distribution; training a preset mathematical model for sensitive data identification until the model converges; and performing threshold calculation on the first tax payment data by using the converged preset mathematical model to obtain a sensitive data identification threshold, and taking data of which the numerical value is greater than the sensitive data identification threshold in the second tax payment data as enterprise tax payment sensitive data, thereby realizing enterprise tax payment sensitive data identification, and accurately identifying and effectively managing the enterprise tax payment sensitive data. And enterprise data leakage is prevented.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, equipment and medium for identifying enterprise tax sensitive data. Background Art

[0002] Tax-sensitive data is extremely critical information in corporate operations. It not only covers the company's financial status, but also directly relates to the company's tax compliance and reputation. Tax authorities calculate and collect taxes based on the data provided by the company. Any inaccurate data may lead to tax audits, fines, or even criminal liability for the company. Tax-sensitive data often contains key information such as the company's business strategy, cost structure, and market positioning. Once this information is obtained by competitors, it may put the company at a disadvantage in the market.

[0003] From the above, it can be seen that how to accurately identify corporate tax sensitive data, prevent corporate data leakage, and effectively manage and protect corporate tax sensitive data are issues to be resolved in this field. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for identifying sensitive data of corporate tax payment, which can accurately identify sensitive data of corporate tax payment, prevent corporate data leakage, and effectively manage and protect sensitive data of corporate tax payment. The specific scheme is as follows:

[0005] In a first aspect, the present application discloses a method for identifying sensitive data of corporate tax payment, comprising:

[0006] Acquire tax data in enterprise operation, and perform data segmentation on the tax data to obtain first tax data for model training and second tax data to be identified;

[0007] Using a diffusion time estimation network to clean and process the first tax data, and construct a diffusion link, input the processed tax data into the diffusion link for calculation to output a vector, and perform posterior distribution calculation on the vector to obtain the posterior distribution of the current time;

[0008] Training a preset mathematical model for sensitive data identification based on the posterior distribution until the preset mathematical model converges to obtain the converged preset mathematical model;

[0009] The first tax data is threshold calculated using the converged preset mathematical model to obtain a sensitive data identification threshold, and the data in the second tax data whose value is greater than the sensitive data identification threshold is used as corporate tax sensitive data to achieve corporate tax sensitive data identification.

[0010] Optionally, the tax data in the business operation includes value-added tax, corporate income tax, financial statements, tax return data, invoice data, other taxes and fees, and tax investigation data.

[0011] Optionally, dividing the tax data includes:

[0012] Determine the data partition ratio based on business needs;

[0013] The tax data is divided according to the data division ratio.

[0014] Optionally, the step of constructing a diffusion link and inputting the processed tax data into the diffusion link for calculation to output a vector includes:

[0015] Constructing a forward diffusion link and a backward diffusion link; the diffusion link includes the forward diffusion link and the backward diffusion link;

[0016] The processed tax data is input into the forward diffusion link for forward diffusion, and then the data after forward diffusion is input into the backward diffusion link, and is calculated using a preset backward diffusion calculation formula to obtain a vector.

[0017] Optionally, the back diffusion calculation formula is:

[0018] ;

[0019] in, is the probability that all sample state transitions from time step 0 to time step T occur simultaneously, is the starting point of the joint probability, , is a Gaussian distribution, and the random variable is , the mean is , is the data of the previous time step, is the data of the current time step, t is the current time step, are learnable parameters, is the variance, and I is the identity matrix with the same dimension as the data.

[0020] Optionally, before training the preset mathematical model for sensitive data identification based on the posterior distribution, the method further includes:

[0021] Constructing a preset mathematical model based on the backward diffusion calculation formula, the Gaussian probability density function formula and the maximum function;

[0022] The Gaussian probability density function formula is:

[0023] ;

[0024] in, is the vector of diffusion link output, is the variance, d is The number of real numbers, D is the data set, for The square of the second normal form of;

[0025] The maximum function is:

[0026] ;

[0027] in, is roughly proportional.

[0028] Optionally, the using the converged preset mathematical model to perform threshold calculation on the first tax data to obtain a sensitive data identification threshold, and taking the data in the second tax data whose value is greater than the sensitive data identification threshold as corporate tax sensitive data, includes:

[0029] Performing threshold calculation on the first tax data using the converged preset mathematical model to obtain a mode or an average corresponding to the first tax data; the sensitive data identification threshold is the mode or the average;

[0030] comparing the mode or the average with all values ​​in the second tax data;

[0031] Data with a value greater than the mode or the average is regarded as enterprise tax sensitive data, and data with a value not greater than the mode or the average is regarded as enterprise tax non-sensitive data.

[0032] In a second aspect, the present application discloses a device for identifying sensitive data of corporate tax payment, comprising:

[0033] A data acquisition and segmentation module, used to acquire tax data in enterprise operations and perform data segmentation on the tax data to obtain first tax data for model training and second tax data to be identified;

[0034] a calculation module, configured to clean and process the first tax data using a diffusion time estimation network, and construct a diffusion link, input the processed tax data into the diffusion link for calculation to output a vector, and perform posterior distribution calculation on the vector to obtain a posterior distribution of the current time;

[0035] A model training module, used for training a preset mathematical model for sensitive data identification based on the posterior distribution until the preset mathematical model converges to obtain the converged preset mathematical model;

[0036] The identification module is used to perform threshold calculation on the first tax data using the preset mathematical model after convergence to obtain a sensitive data identification threshold, and to use the data in the second tax data whose value is greater than the sensitive data identification threshold as corporate tax sensitive data to achieve corporate tax sensitive data identification.

[0037] In a third aspect, the present application discloses an electronic device, comprising:

[0038] Memory, used to store computer programs;

[0039] The processor is used to execute the computer program to implement the aforementioned method for identifying corporate tax sensitive data.

[0040] In a fourth aspect, the present application discloses a computer storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the aforementioned disclosed method for identifying corporate tax sensitive data are implemented.

[0041] It can be seen that the present application provides a method for identifying sensitive data of enterprise tax payment, including obtaining tax data in enterprise operation, dividing the tax data to obtain first tax data for model training and second tax data to be identified; using a diffusion time estimation network to clean and process the first tax data, and constructing a diffusion link, inputting the processed tax data into the diffusion link for calculation to output a vector, and performing posterior distribution calculation on the vector to obtain the posterior distribution of the current time; training a preset mathematical model for sensitive data identification based on the posterior distribution until the preset mathematical model converges to obtain the preset mathematical model after convergence; using the preset mathematical model after convergence to perform threshold calculation on the first tax data to obtain a sensitive data identification threshold, and taking the data in the second tax data whose value is greater than the sensitive data identification threshold as enterprise tax sensitive data to realize enterprise tax sensitive data identification. The present application first obtains the tax data in the enterprise operation, divides the tax data into the first tax data for model training and the second tax data to be identified, solves the problem of data annotation required in the prior art, then cleans and processes the first tax data to improve the accuracy of the data, builds a diffusion link, inputs the processed tax data into the diffusion link for calculation, outputs a vector, calculates the posterior distribution of the vector, obtains the posterior distribution of the current time, trains the preset mathematical model until the preset mathematical model converges, uses the converged preset mathematical model to calculate the threshold of the first tax data, obtains the sensitive data recognition threshold, and uses the data in the second tax data whose value is greater than the sensitive data recognition threshold as the enterprise tax sensitive data. Through self-supervision tasks, good representations are learned, the unknown parameters of each variable are estimated, and the overall model is improved. The enterprise tax sensitive data can be accurately identified, the enterprise data leakage is prevented, and the enterprise tax sensitive data is effectively managed and protected. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0043] Figure 1 A flow chart of a method for identifying sensitive data of corporate tax payment disclosed in this application;

[0044] Figure 2 A specific flow chart for identifying sensitive data of corporate tax payment disclosed in this application;

[0045] Figure 3This is a schematic diagram of the structure of a device for identifying sensitive data of corporate tax payment disclosed in this application;

[0046] Figure 4 A structural diagram of an electronic device provided for this application. DETAILED DESCRIPTION

[0047] 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.

[0048] Tax-sensitive data is extremely critical information in corporate operations. It not only covers the company's financial status, but is also directly related to the company's tax compliance and reputation. Tax authorities calculate and collect taxes based on the data provided by the company. Any inaccurate data may cause the company to face tax audits, fines, or even criminal liability. Tax-sensitive data often contains key information such as the company's business strategy, cost structure, and market positioning. Once this information is obtained by competitors, it may lead to a disadvantageous position for the company in the market. As can be seen from the above, how to accurately identify corporate tax-sensitive data, prevent corporate data leakage, and effectively manage and protect corporate tax-sensitive data are issues to be solved in this field.

[0049] See also Figure 1 As shown, the embodiment of the present invention discloses a method for identifying sensitive data of corporate tax payment, which may specifically include:

[0050] Step S11: Acquire tax data in enterprise operations, and perform data segmentation on the tax data to obtain first tax data for model training and second tax data to be identified.

[0051] In this embodiment, tax data in enterprise operations is obtained; the tax data in enterprise operations includes value-added tax, corporate income tax, financial statements, tax declaration data, invoice data, other taxes and fees, and tax investigation data; then the data division ratio is determined according to business needs, and the tax data is divided according to the data division ratio to obtain the first tax data for model training and the second tax data to be identified.

[0052] Specifically, VAT (Value-added Tax) is the tax collected by enterprises in the process of selling goods or providing services, including VAT declaration information, such as the type of taxable vouchers, declaration form information, etc.; Corporate income tax is the tax that enterprises need to pay on their net profit in a certain period of time, which is usually reflected in the income statement of financial statements as part of taxes and surcharges; Financial statements include balance sheets, income statements, cash flow statements, etc. These statements contain the financial status of the enterprise and indirectly reflect the tax situation of the enterprise. For example, the tax payable item in the balance sheet can reflect the taxes payable by the enterprise; Tax declaration data is the relevant information filled in by taxpayers on the declaration form. Other tax collection and management system data other than customs data and declaration data, including taxpayers' basic information, production and operation information (such as invoices, etc.); invoice data: the invoice date, quantity, unit price, amount, tax amount, company name and other information of general value-added tax invoices. These data will be stored and transmitted to the tax big data cloud platform of the State Administration of Taxation; other taxes and fees: such as urban construction tax, education surcharge, local education surcharge, etc., which usually appear as taxes and additional items in the income statement; tax survey data: tax statistics and survey data conducted by tax authorities, which help to understand the tax compliance and tax burden of enterprises.

[0053] Step S12: Use a diffusion time estimation network to clean and process the first tax data, and build a diffusion link, input the processed tax data into the diffusion link for calculation to output a vector, and perform posterior distribution calculation on the vector to obtain the posterior distribution of the current time.

[0054] In this embodiment, a DTE (Diffusion Time Estimation) network is used to clean and process the first tax data, and a forward diffusion link and a backward diffusion link are constructed; the diffusion link includes the forward diffusion link and the backward diffusion link, the processed tax data is input into the forward diffusion link for forward diffusion, and then the forward diffusion data is input into the backward diffusion link, and a preset backward diffusion calculation formula is used for calculation to obtain a vector, and a posterior distribution calculation is performed on the vector to obtain a posterior distribution of the current time;

[0055] Wherein, the backward diffusion calculation formula is:

[0056] ;

[0057] in, is the probability that all sample state transitions from time step 0 to time step T occur simultaneously, is the starting point of the joint probability, , is a Gaussian distribution, and the random variable is , the mean is , is the data of the previous time step, is the data of the current time step, t is the current time step, are learnable parameters, is the variance, and I is the identity matrix with the same dimension as the data.

[0058] Specifically, the data collected in the first step are cleaned and processed to obtain the required data set, and each sample data is represented by a symbol Next, we use mathematical language to describe the problem to be solved. express from Here is a sample drawn from represents sample data, Represents the probability distribution of sample data; Represents the corresponding hidden variables, which are unobservable. They represent the states of the sample data at each time step in the diffusion process, making it easier for the model to understand the generation process of the entire diffusion link; Represents the variance series, used for forward diffusion link calculation.

[0059] Among them, the forward diffusion link calculation process is:

[0060] In this stage, the model gradually adds noise to the data until the data is completely transformed into a noise distribution. Each step in this process involves a small change in the data, which ultimately leads to the loss of the original data signal. To complete this process, it is necessary to define an approximate posterior distribution, which represents the model's attempt to approximate the estimate of the true posterior distribution, using express:

[0061] ;

[0062] ;

[0063] in, represents a Gaussian distribution whose random variable is , the mean is , the variance is is the variance at time step t, and I is the identity matrix with the same dimension as the data;

[0064] In the forward diffusion link, each step involves transferring the current data Transformed to the next time step in some way , which is usually achieved by adding noise, and the noise comes from sampling from a Gaussian distribution. The mathematical expression is defined as follows:

[0065] ;

[0066] in, ,and , represents the variance portion of the data retained in the tth step, that is, , Indicates all The product of

[0067] The calculation process of the backward diffusion link is:

[0068] The subsequent diffusion link is the reverse process of the forward diffusion. The model starts with pure noise, gradually removes the noise, and finally reconstructs the original data. In the backward diffusion process, the probability of all sample state transitions occurring simultaneously from time step 0 to T is called the joint probability. The mathematical definition of the backward diffusion link based on this joint probability is the above backward diffusion calculation formula.

[0069] Step S13: training a preset mathematical model for sensitive data identification based on the posterior distribution until the preset mathematical model converges to obtain the converged preset mathematical model.

[0070] In this embodiment, a preset mathematical model is constructed based on the backward diffusion calculation formula, the Gaussian probability density function formula and the maximum function, and the preset mathematical model for sensitive data identification is trained based on the posterior distribution until the preset mathematical model converges to obtain the converged preset mathematical model;

[0071] The Gaussian probability density function formula is:

[0072] ;

[0073] in, is the vector of diffusion link output, is the variance, d is The number of real numbers, D is the data set, for The square of the second normal form of;

[0074] The maximum function is:

[0075] ;

[0076] in, is roughly proportional.

[0077] In this embodiment, in the above diffusion link, ( is a vector of d real numbers), and at time t, let the variance be , the data set is D, the data distribution is p(x), according to Bayes' theorem, the posterior distribution is proportional to the product of the prior distribution and the likelihood function, the mathematical expression is as follows:

[0078] ;

[0079] The above Gaussian distribution is replaced by a Gaussian probability density function to obtain the Gaussian probability density function formula.

[0080] Step S14: Perform threshold calculation on the first tax data using the converged preset mathematical model to obtain a sensitive data identification threshold, and use the data in the second tax data whose value is greater than the sensitive data identification threshold as corporate tax sensitive data to achieve corporate tax sensitive data identification.

[0081] In this embodiment, a preset mathematical model after convergence is used to perform threshold calculation on the first tax data to obtain the mode or mean corresponding to the first tax data; the sensitive data identification threshold is the mode or the mean; the mode or the mean is compared with all values ​​in the second tax data; data with values ​​greater than the mode or the mean are used as corporate tax sensitive data, and data with values ​​not greater than the mode or the mean are used as corporate tax non-sensitive data, so as to realize corporate tax sensitive data identification.

[0082] The specific process of this application to realize the identification of sensitive data of corporate tax payment is as follows: Figure 2 As shown, first, the tax data of the enterprise is obtained and the tax data is divided; then the tax data is cleaned and processed using the DTE network, and a forward diffusion link and a backward diffusion link are constructed, and the processed tax data is input into the forward diffusion link and the backward diffusion link for calculation to output a vector, and the posterior distribution of the vector is calculated to obtain the posterior distribution of the current time; then, based on the posterior distribution, the preset mathematical model for sensitive data identification is trained until the preset mathematical model converges to obtain the converged preset mathematical model; finally, the converged preset mathematical model is used to identify the sensitive data of the enterprise tax.

[0083] Since sensitive tax data in the tax data of enterprises usually accounts for only a small part of the overall data, this situation is not very friendly to supervised learning methods because the data annotation stage may be difficult. However, the DTE network used in this application is an unsupervised learning method that can effectively solve this problem because it does not require data annotation. Secondly, after receiving the data, the model will build a diffusion chain for the data, including a forward diffusion chain and a backward diffusion chain; based on the vector data generated during the diffusion chain construction process, its posterior distribution is calculated through its likelihood function and prior distribution; then, through further mathematical derivation and simplified operations, the final posterior distribution expression is obtained; finally, based on the posterior distribution expression, its mode and mean are calculated as the basis for dividing sensitive data.

[0084] The key points of this application are: constructing a DTE network and using mathematical language to describe the problem; the data processed by the algorithm are all from tax big data, such as: personal basic information, income information, special deduction information, special additional deduction information, donation information, overseas income information, tax return information, and tax payment certificates, in order to improve the interpretability of the model in the tax direction; through self-supervision tasks to learn good representations, estimate the unknown parameters of each variable, and improve the overall model.

[0085] In this embodiment, the tax data in the enterprise operation is obtained, and the tax data is divided to obtain the first tax data for model training and the second tax data to be identified; the first tax data is cleaned and processed by using the diffusion time estimation network, and a diffusion link is constructed, the processed tax data is input into the diffusion link for calculation to output a vector, and the posterior distribution of the vector is calculated to obtain the posterior distribution of the current time; the preset mathematical model for sensitive data identification is trained based on the posterior distribution until the preset mathematical model converges to obtain the preset mathematical model after convergence; the threshold of the first tax data is calculated by using the preset mathematical model after convergence to obtain the sensitive data identification threshold, and the data in the second tax data whose value is greater than the sensitive data identification threshold is used as the enterprise tax sensitive data to realize the enterprise tax sensitive data identification. The present application first obtains the tax data in the enterprise operation, divides the tax data into the first tax data for model training and the second tax data to be identified, solves the problem of data annotation required in the prior art, then cleans and processes the first tax data to improve the accuracy of the data, builds a diffusion link, inputs the processed tax data into the diffusion link for calculation, outputs a vector, calculates the posterior distribution of the vector, obtains the posterior distribution of the current time, trains the preset mathematical model until the preset mathematical model converges, uses the converged preset mathematical model to calculate the threshold of the first tax data, obtains the sensitive data recognition threshold, and uses the data in the second tax data whose value is greater than the sensitive data recognition threshold as the enterprise tax sensitive data. Through self-supervision tasks, good representations are learned, the unknown parameters of each variable are estimated, and the overall model is improved. The enterprise tax sensitive data can be accurately identified, the enterprise data leakage is prevented, and the enterprise tax sensitive data is effectively managed and protected.

[0086] See also Figure 3 As shown, the embodiment of the present invention discloses a device for identifying sensitive data of enterprise tax payment, which may specifically include:

[0087] The data acquisition and segmentation module 11 is used to acquire the tax data in the enterprise operation and perform data segmentation on the tax data to obtain the first tax data for model training and the second tax data to be identified;

[0088] A calculation module 12, configured to clean and process the first tax data using a diffusion time estimation network, and construct a diffusion link, input the processed tax data into the diffusion link for calculation to output a vector, and perform posterior distribution calculation on the vector to obtain a posterior distribution of the current time;

[0089] A model training module 13, used for training a preset mathematical model for sensitive data identification based on the posterior distribution until the preset mathematical model converges to obtain the converged preset mathematical model;

[0090] The identification module 14 is used to perform threshold calculation on the first tax data using the preset mathematical model after convergence to obtain a sensitive data identification threshold, and to use the data in the second tax data whose value is greater than the sensitive data identification threshold as corporate tax sensitive data to achieve corporate tax sensitive data identification.

[0091] In this embodiment, the tax data in the enterprise operation is obtained, and the tax data is divided to obtain the first tax data for model training and the second tax data to be identified; the first tax data is cleaned and processed by using the diffusion time estimation network, and a diffusion link is constructed, the processed tax data is input into the diffusion link for calculation to output a vector, and the posterior distribution of the vector is calculated to obtain the posterior distribution of the current time; the preset mathematical model for sensitive data identification is trained based on the posterior distribution until the preset mathematical model converges to obtain the preset mathematical model after convergence; the threshold of the first tax data is calculated by using the preset mathematical model after convergence to obtain the sensitive data identification threshold, and the data in the second tax data whose value is greater than the sensitive data identification threshold is used as the enterprise tax sensitive data to realize the enterprise tax sensitive data identification. The present application first obtains the tax data in the enterprise operation, divides the tax data into the first tax data for model training and the second tax data to be identified, solves the problem of data annotation required in the prior art, then cleans and processes the first tax data to improve the accuracy of the data, builds a diffusion link, inputs the processed tax data into the diffusion link for calculation, outputs a vector, calculates the posterior distribution of the vector, obtains the posterior distribution of the current time, trains the preset mathematical model until the preset mathematical model converges, uses the converged preset mathematical model to calculate the threshold of the first tax data, obtains the sensitive data recognition threshold, and uses the data in the second tax data whose value is greater than the sensitive data recognition threshold as the enterprise tax sensitive data. Through self-supervision tasks, good representations are learned, the unknown parameters of each variable are estimated, and the overall model is improved. The enterprise tax sensitive data can be accurately identified, the enterprise data leakage is prevented, and the enterprise tax sensitive data is effectively managed and protected.

[0092] In some specific embodiments, the tax data in the business operation includes value-added tax, corporate income tax, financial statements, tax return data, invoice data, other taxes and fees, and tax investigation data.

[0093] In some specific embodiments, the data acquisition and division module 11 may specifically include:

[0094] A partition ratio determination module is used to determine the data partition ratio according to business requirements;

[0095] A data partitioning module is used to partition the tax data according to the data partitioning ratio.

[0096] In some specific embodiments, the calculation module 12 may specifically include:

[0097] A diffusion link construction module, used to construct a forward diffusion link and a backward diffusion link; the diffusion link includes the forward diffusion link and the backward diffusion link;

[0098] The input module is used to input the processed tax data into the forward diffusion link for forward diffusion, and then input the data after forward diffusion into the backward diffusion link, and use the preset backward diffusion calculation formula for calculation to obtain a vector.

[0099] In some specific embodiments, the back diffusion calculation formula is:

[0100] ;

[0101] in, is the probability that all sample state transitions from time step 0 to time step T occur simultaneously, is the starting point of the joint probability, , is a Gaussian distribution, and the random variable is , the mean is , is the data of the previous time step, is the data of the current time step, t is the current time step, are learnable parameters, is the variance, and I is the identity matrix with the same dimension as the data.

[0102] In some specific embodiments, the model training module 13 may specifically include:

[0103] A mathematical model building module, used to build a preset mathematical model based on the backward diffusion calculation formula, the Gaussian probability density function formula and the maximum function;

[0104] The Gaussian probability density function formula is:

[0105] ;

[0106] in, is the vector of diffusion link output, is the variance, d is The number of real numbers, D is the data set, for The square of the second normal form of;

[0107] The maximum function is:

[0108] ;

[0109] in, is roughly proportional.

[0110] In some specific embodiments, the identification module 14 may specifically include:

[0111] A threshold calculation module, used to perform threshold calculation on the first tax data using a converged preset mathematical model to obtain a mode or an average corresponding to the first tax data; the sensitive data identification threshold is the mode or the average;

[0112] A comparison module, used for comparing the mode or the average with all the values ​​in the second tax data;

[0113] The data determination module is used to treat data with a value greater than the mode or the average as enterprise tax sensitive data, and treat data with a value not greater than the mode or the average as enterprise tax non-sensitive data.

[0114] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the method for identifying sensitive corporate tax data performed by the electronic device disclosed in any of the aforementioned embodiments.

[0115] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0116] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon include an operating system 221, a computer program 222 and data 223, etc. The storage method can be temporary storage or permanent storage.

[0117] Among them, the operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20 to realize the operation and processing of the data 223 in the memory 22 by the processor 21, which can be Windows, Unix, Linux, etc. In addition to including a computer program that can be used to complete the enterprise tax sensitive data identification method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks. In addition to data received by the enterprise tax sensitive data identification device and transmitted from an external device, the data 223 can also include data collected by its own input and output interface 25, etc.

[0118] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0119] Furthermore, an embodiment of the present application also discloses a computer-readable storage medium, in which a computer program is stored. When the computer program is loaded and executed by a processor, the steps of the method for identifying corporate tax sensitive data disclosed in any of the aforementioned embodiments are implemented.

[0120] Finally, 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. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0121] The above is a detailed introduction to a method, device, equipment and storage medium for identifying enterprise tax sensitive data provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for identifying sensitive data of corporate tax payment, characterized in that: include: Acquire tax data in enterprise operation, and perform data segmentation on the tax data to obtain first tax data for model training and second tax data to be identified; Using a diffusion time estimation network to clean and process the first tax data, and construct a diffusion link, input the processed tax data into the diffusion link for calculation to output a vector, and perform posterior distribution calculation on the vector to obtain the posterior distribution of the current time; Training a preset mathematical model for sensitive data identification based on the posterior distribution until the preset mathematical model converges to obtain the converged preset mathematical model; The first tax data is threshold calculated using the converged preset mathematical model to obtain a sensitive data identification threshold, and the data in the second tax data whose value is greater than the sensitive data identification threshold is used as corporate tax sensitive data to achieve corporate tax sensitive data identification.

2. The method for identifying sensitive data of corporate tax payment according to claim 1, characterized in that: The tax data in the business operations include value added tax, corporate income tax, financial statements, tax return data, invoice data, other taxes and fees, and tax investigation data.

3. The method for identifying sensitive data of corporate tax payment according to claim 1, characterized in that: The step of dividing the tax data includes: Determine the data partition ratio based on business needs; The tax data is divided according to the data division ratio.

4. The method for identifying sensitive data of corporate tax payment according to claim 1, characterized in that: The step of constructing a diffusion link and inputting the processed tax data into the diffusion link for calculation to output a vector includes: Constructing a forward diffusion link and a backward diffusion link; the diffusion link includes the forward diffusion link and the backward diffusion link; The processed tax data is input into the forward diffusion link for forward diffusion, and then the data after forward diffusion is input into the backward diffusion link, and is calculated using a preset backward diffusion calculation formula to obtain a vector.

5. The method for identifying sensitive data of corporate tax payment according to claim 4, characterized in that: The back diffusion calculation formula is: ； in, is the probability that all sample state transitions from time step 0 to time step T occur simultaneously, is the starting point of the joint probability, , is a Gaussian distribution, and the random variable is , the mean is , is the data of the previous time step, is the data of the current time step, t is the current time step, are learnable parameters, is the variance, and I is the identity matrix with the same dimension as the data.

6. The method for identifying sensitive data of corporate tax payment according to claim 5, characterized in that: Before training the preset mathematical model for sensitive data identification based on the posterior distribution, the method further includes: Constructing a preset mathematical model based on the backward diffusion calculation formula, the Gaussian probability density function formula and the maximum function; The Gaussian probability density function formula is: ； in, is the vector of diffusion link output, is the variance, d is The number of real numbers, D is the data set, for The square of the second normal form of; The maximum function is: ； in, is roughly proportional.

7. The method for identifying sensitive data of corporate tax payment according to any one of claims 1 to 6, characterized in that: The using the converged preset mathematical model to perform threshold calculation on the first tax data to obtain a sensitive data identification threshold, and taking the data in the second tax data whose value is greater than the sensitive data identification threshold as enterprise tax sensitive data, includes: Performing threshold calculation on the first tax data using the converged preset mathematical model to obtain a mode or an average corresponding to the first tax data; the sensitive data identification threshold is the mode or the average; comparing the mode or the average with all values ​​in the second tax data; Data with a value greater than the mode or the average is regarded as enterprise tax sensitive data, and data with a value not greater than the mode or the average is regarded as enterprise tax non-sensitive data.

8. An enterprise tax sensitive data identification device, characterized in that: include: A data acquisition and segmentation module, used to acquire tax data in enterprise operations and perform data segmentation on the tax data to obtain first tax data for model training and second tax data to be identified; a calculation module, configured to clean and process the first tax data using a diffusion time estimation network, and construct a diffusion link, input the processed tax data into the diffusion link for calculation to output a vector, and perform posterior distribution calculation on the vector to obtain a posterior distribution of the current time; A model training module, used for training a preset mathematical model for sensitive data identification based on the posterior distribution until the preset mathematical model converges to obtain the converged preset mathematical model; The identification module is used to perform threshold calculation on the first tax data using the preset mathematical model after convergence to obtain a sensitive data identification threshold, and to use the data in the second tax data whose value is greater than the sensitive data identification threshold as corporate tax sensitive data to achieve corporate tax sensitive data identification.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the method for identifying corporate tax sensitive data as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store computer programs; wherein, when the computer program is executed by a processor, the method for identifying corporate tax sensitive data as described in any one of claims 1 to 7 is implemented.

Images