Financial service platform based on privacy computing

Abstract

The embodiment of the invention relates to the technical field of financial services, and particularly discloses a financial service platform based on privacy computing, which comprises a service request processing unit used for receiving a financial service request and determining a service request object and a financial service business; the data collecting and processing unit is used for acquiring a plurality of pieces of related collected data and carrying out standardization and range desensitization processing on the plurality of pieces of related collected data; the model training and verification unit is used for selecting a basic privacy calculation model and carrying out model training and verification; a model deployment application unit; and a model iteration updating unit. According to the method, standardization and range desensitization processing can be carried out on a plurality of related collected data, a basic privacy calculation model is selected, model training and verification are carried out, and then model deployment application and model iteration updating are carried out, so that an effective technical standard is provided, a complex and tedious technical process is avoided, desensitization processing does not need to be carried out on all privacy information, and the efficiency is improved. The computing resources are effectively reduced, and the time cost is shortened.

Description

Technical Field

[0001] The present invention belongs to the field of financial service technology, and in particular to a financial service platform based on privacy computing. Background Art

[0002] Privacy computing refers to an information technology system that implements data analysis and data value mining under the premise of protecting the data itself from external leakage. It is a cross-integration of multiple technologies such as artificial intelligence, cryptography, and data science.

[0003] Financial services based on privacy computing are an important development direction in the field of financial technology in recent years. They aim to resolve the contradiction between data circulation and privacy protection in financial services.

[0004] In the existing technology, financial services based on privacy computing is an emerging technology field. It lacks effective technical standards and the implementation process is relatively complicated and cumbersome. In addition, during the privacy desensitization process, all private information needs to be desensitized, which requires huge computing resources and time costs. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a financial service platform based on privacy computing, aiming to solve the problems raised in the background technology.

[0006] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions: A financial service platform based on privacy computing, the platform includes a service request processing unit, a data collection processing unit, a model training and verification unit, a model deployment and application unit, and a model iteration and update unit, wherein: A service request processing unit, configured to receive a financial service request, analyze the financial service request, determine a service request object and a financial service business, and generate and sign a financial service agreement with the service request object; A data collection and processing unit, configured to collect data according to the financial service agreement, obtain a plurality of relevant collected data, and perform standardization and range desensitization processing on the plurality of relevant collected data to generate a plurality of target collected data; A model training and verification unit, used to select a basic privacy computing model, collect data through multiple targets, perform model training and verification on the basic privacy computing model, and generate a target privacy computing model; A model deployment application unit, used to deploy the target privacy computing model in the financial service business, and to perform application monitoring and adjustment optimization on the target privacy computing model; The model iteration and updating unit is used to continuously collect multiple related process data in the financial service business, iterate the target privacy computing model through the multiple related process data, and update the iterative privacy computing model.

[0007] As a further limitation of the technical solution of the embodiment of the present invention, the service request processing unit specifically includes: A service request receiving module, used for receiving a financial service request; A service request analysis module, used to analyze the financial service request and determine the service request object and the financial service business; A service agreement generation module, used to generate a financial service agreement; A service agreement signing module is used to sign the financial service agreement with the service request object.

[0008] As a further limitation of the technical solution of the embodiment of the present invention, the service agreement generation module specifically includes: A service analysis submodule, for clarifying service responsibilities, service rights and service scope according to the financial service business; A service terms formulation submodule is used to formulate relevant service terms according to the service responsibilities, service rights and service scope; The legal and regulatory matching submodule is used to match relevant laws and regulations; The service agreement creation submodule is used to integrate the relevant service terms and the relevant laws and regulations to create a financial service agreement.

[0009] As a further limitation of the technical solution of the embodiment of the present invention, the data collection and processing unit specifically includes: A related type determination module, used to determine a plurality of related data types according to the financial service agreement; A related data collection module, used to collect data according to a plurality of related data types, and obtain a plurality of related collected data; A standardization processing module, used for performing standardization processing on the plurality of related collected data to obtain a plurality of standard collected data; The range desensitization processing module is used to perform range desensitization processing on the multiple standard collection data to generate multiple target collection data.

[0010] As a further limitation of the technical solution of the embodiment of the present invention, the range desensitization processing module specifically includes: An object identity range determination submodule, used to determine the object identity range of the service request object; A privacy identification submodule, used to perform privacy identification on a plurality of the standard collected data and mark a plurality of privacy identity information; A privacy identity range identification submodule, used to identify the privacy identity ranges of the plurality of privacy identity information; A range comparison submodule, configured to compare a plurality of the privacy identity ranges with the object identity range; A retention processing submodule, used for retaining the privacy identity information corresponding to the multiple privacy identity ranges larger than the object identity range; The replacement desensitization submodule is used to replace and desensitize the privacy identity information corresponding to multiple privacy identity ranges that are not larger than the object identity range.

[0011] As a further limitation of the technical solution of the embodiment of the present invention, the model training verification unit specifically includes: A basic model selection module, used to select a basic privacy computing model according to the financial service business; A collection data division module, used for dividing the plurality of target collection data into a test set and a validation set; The model training and verification module is used to perform model training and verification on the basic privacy computing model through the test set and the verification set to generate a target privacy computing model.

[0012] As a further limitation of the technical solution of the embodiment of the present invention, the data collection and division module specifically includes: The ratio acquisition submodule is used to obtain the data division ratio; An instruction generation submodule, used for generating a data partitioning instruction according to the data partitioning ratio; The partitioning processing submodule is used to partition the data collected from the plurality of targets according to the data partitioning instruction to obtain a test set and a verification set.

[0013] As a further limitation of the technical solution of the embodiment of the present invention, the basic privacy computing model specifically includes: a multi-party secure computing model, a homomorphic encryption model or a differential privacy model.

[0014] As a further limitation of the technical solution of the embodiment of the present invention, the model deployment application unit specifically includes: A model deployment module, used to deploy the target privacy computing model in the financial service business; An application monitoring module, used to perform application monitoring on the target privacy computing model and record application monitoring data including computing accuracy, response time and response anomalies; An adjustment and optimization module is used to adjust and optimize the target privacy computing model according to the application monitoring data.

[0015] As a further limitation of the technical solution of the embodiment of the present invention, the model iterative updating unit specifically includes: A continuous collection module, used for continuously collecting a plurality of relevant process data in the financial service business; A data screening module, used for screening a plurality of valid process data from a plurality of the related process data; The model iteration module is used to iterate the target privacy computing model through multiple valid process data to update the iterative privacy computing model.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The financial service platform based on privacy computing disclosed in the embodiment of the present invention includes: a service request processing unit, which is used to receive financial service requests, determine the service request object and the financial service business; a data collection processing unit, which is used to obtain multiple related collected data, and standardize and scope-desensitize the multiple related collected data; a model training and verification unit, which is used to select a basic privacy computing model, perform model training and verification; a model deployment and application unit; and a model iteration and update unit. It is capable of standardizing and scope-desensitizing multiple related collected data, selecting a basic privacy computing model, performing model training and verification, and then performing model deployment and application and model iteration and update, providing effective technical standards, avoiding complex and cumbersome technical processes, and not requiring all privacy information to be desensitized, effectively reducing computing resources and shortening time costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0018] Figure 1 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0019] Figure 2 The structure block diagram of the service request processing unit in the system provided by the embodiment of the present invention is shown.

[0020] Figure 3 The structure block diagram of the service protocol generation module in the system provided by the embodiment of the present invention is shown.

[0021] Figure 4 The structure block diagram of the data collection and processing unit in the system provided by the embodiment of the present invention is shown.

[0022] Figure 5 A structural block diagram of a range desensitization processing module in a system provided by an embodiment of the present invention is shown.

[0023] Figure 6 A structural block diagram of a model training and verification unit in a system provided by an embodiment of the present invention is shown.

[0024] Figure 7 The structure block diagram of the data collection and partitioning module in the system provided by the embodiment of the present invention is shown.

[0025] Figure 8 A structural block diagram of a model deployment application unit in a system provided by an embodiment of the present invention is shown.

[0026] Fig. 9 A structural block diagram of a model iterative updating unit in a system provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0028] It is understandable that financial services based on privacy computing are an important development direction in the field of financial technology in recent years. It aims to resolve the contradiction between data circulation and privacy protection in financial services. In the existing technology, financial services based on privacy computing are an emerging technical field, lacking effective technical standards, and the implementation process is relatively complicated and cumbersome. In the privacy desensitization process, all private information needs to be desensitized, which requires huge computing resources and time costs.

[0029] To solve the above problems, the financial service platform based on privacy computing disclosed in the embodiment of the present invention includes: a service request processing unit, which is used to receive financial service requests, analyze the financial service requests, determine the service request object and the financial service business, generate and sign a financial service agreement with the service request object; a data collection processing unit, which is used to collect data according to the financial service agreement, obtain multiple related collected data, and standardize and range-massify the multiple related collected data to generate multiple target collected data; a model training and verification unit, which is used to select a basic privacy computing model, collect data through multiple targets, train and verify the basic privacy computing model, and generate a target privacy computing model; a model deployment and application unit, which is used to deploy the target privacy computing model in the financial service business, and monitor and adjust and optimize the target privacy computing model; a model iteration and update unit, which is used to continuously collect multiple related process data in the financial service business, iterate the target privacy computing model through multiple related process data, and update the iterative privacy computing model. It can standardize and scope-desensitize multiple related collected data, select basic privacy computing models, conduct model training and verification, and then deploy and apply the models and iterate and update the models, providing effective technical standards and avoiding complex and tedious technical processes. There is no need to desensitize all privacy information, which effectively reduces computing resources and shortens time costs.

[0030] Figure 1 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0031] Specifically, the financial service platform based on privacy computing includes: The service request processing unit 101 is used to receive a financial service request, analyze the financial service request, determine a service request object and a financial service business, and generate and sign a financial service agreement with the service request object.

[0032] In an embodiment of the present invention, the service request processing unit 101 receives a financial service request, then identifies the content of the financial service request, determines the service request object and the financial service business, clarifies the service responsibilities, service rights and service scope according to the financial service business, formulates relevant service terms in accordance with the service responsibilities, service rights and service scope, matches the relevant laws and regulations of the financial service business, integrates the relevant service terms and relevant laws and regulations, creates a financial service agreement, and then signs the financial service agreement with the service request object.

[0033] It is understandable that the service request object can be an e-commerce company, a game company, a bank, a network operator, etc.

[0034] Specifically, Figure 2It shows a structural block diagram of the service request processing unit 101 in the system provided by the embodiment of the present invention.

[0035] In the preferred implementation of the present invention, the service request processing unit 101 specifically includes: The service request receiving module 1011 is used to receive a financial service request.

[0036] The service request analysis module 1012 is used to analyze the financial service request and determine the service request object and the financial service business.

[0037] The service agreement generation module 1013 is used to generate a financial service agreement.

[0038] Specifically, Figure 3 It shows a structural block diagram of the service agreement generation module 1013 in the system provided by the embodiment of the present invention.

[0039] Among them, in the preferred implementation mode provided by the present invention, the service agreement generation module 1013 specifically includes: The service analysis submodule 10131 is used to clarify the service responsibilities, service rights and service scope according to the financial service business.

[0040] The service terms formulation submodule 10132 is used to formulate relevant service terms according to the service responsibilities, the service rights and the service scope.

[0041] The law and regulation matching submodule 10133 is used to match relevant laws and regulations.

[0042] The service agreement creation submodule 10134 is used to create a financial service agreement by integrating the relevant service terms and the relevant laws and regulations.

[0043] Furthermore, the service request processing unit 101 further includes: The service agreement signing module 1014 is used to sign the financial service agreement with the service request object.

[0044] Furthermore, the financial service platform based on privacy computing also includes: The data collection and processing unit 102 is used to collect data according to the financial service agreement, obtain multiple related collected data, and perform standardization and range desensitization processing on the multiple related collected data to generate multiple target collected data.

[0045] In an embodiment of the present invention, the data collection and processing unit 102 determines multiple related data types according to the financial service agreement, and then collects data according to the multiple related data types to obtain multiple related collected data, and then performs standardization processing such as numerical normalization and category coding on the multiple related collected data according to a preset standardized format to obtain multiple standard collected data, determine the object identity range of the service request object, and perform privacy identification on the multiple standard collected data, mark multiple privacy identity information, and then identify the privacy identity range of the multiple privacy identity information, by comparing the multiple privacy identity ranges with the object identity range, retaining the privacy identity information corresponding to the multiple privacy identity ranges that are larger than the object identity range, and replacing and desensitizing the privacy identity information corresponding to the multiple privacy identity ranges that are not larger than the object identity range.

[0046] It can be understood that the scope desensitization processing is a process of comparing the privacy identity scope with the object identity scope, and retaining the privacy identity information or replacing the privacy identity information with desensitization according to different comparison results. For example: if the object identity scope is "Company A", if the privacy identity scope is "City B", "Company A" is a company belonging to "City B", then the privacy identity scope is larger than the object identity scope. At this time, the corresponding privacy identity information is retained; if the privacy identity scope is "Name C", "C" is the name of an employee in "Company A", then the privacy identity scope is not larger than the object identity scope. At this time, the corresponding privacy identity information is replaced and desensitized by generating replacement information.

[0047] Specifically, Figure 4 It shows a structural block diagram of the data collection and processing unit 102 in the system provided by an embodiment of the present invention.

[0048] In the preferred embodiment of the present invention, the data collection and processing unit 102 specifically includes: The related type determination module 1021 is used to determine a plurality of related data types according to the financial service agreement.

[0049] The relevant data collection module 1022 is used to collect data according to the multiple relevant data types and obtain multiple relevant collected data.

[0050] A standardization processing module 1023 is used to perform standardization processing on the plurality of related collected data to obtain a plurality of standard collected data; Among them, the relationship between the standardization of the plurality of related collected data is: ; in, represents the original data matrix, represents the mean vector of the data, represents the covariance matrix of the data, represents the standardized data matrix, and Both represent constants used to adjust the standardized results. represents a constant used to avoid zero denominator in logarithmic operations. represents element-wise multiplication; Specifically, by subtracting the mean vector and multiplying by the inverse of the covariance matrix, the various features of the data have the same dimension and center, thereby improving the robustness of the model at different feature scales; at the same time, by using the inverse covariance matrix, the data is further standardized, reducing the correlation between features, which is helpful for the calculation of data that assumes independent and identical distribution.

[0051] A range desensitization processing module 1024 is used to perform range desensitization processing on the plurality of standard collection data to generate a plurality of target collection data; Among them, the relationship formula that exists in the process of performing range desensitization processing on multiple standard collection data using the range desensitization processing module is: ; in, represents the data matrix after pseudonymization, represents the mask matrix, represents the result of hashing the data matrix, represents the Hadamard product, Represents a constant used to adjust the hashing result. and Both represent the coefficients used to adjust the cosine function.

[0052] Specifically, the data is selectively desensitized by using a mask matrix and nonlinear transformation is introduced. At the same time, by hashing the data, sensitive data can be protected to prevent data leakage during model processing.

[0053] Specifically, Figure 5 A structural block diagram of the range desensitization processing module 1024 in the system provided in an embodiment of the present invention is shown.

[0054] Among them, in the preferred embodiment provided by the present invention, the range desensitization processing module 1024 specifically includes: The object identity range determination submodule 10241 is used to determine the object identity range of the service request object.

[0055] The privacy identification submodule 10242 is used to perform privacy identification on multiple standard collected data and mark multiple privacy identity information.

[0056] The privacy identity range identification submodule 10243 is used to identify the privacy identity ranges of the multiple privacy identity information.

[0057] The range comparison submodule 10244 is used to compare the multiple privacy identity ranges with the object identity range.

[0058] The retention processing submodule 10245 is used to retain the privacy identity information corresponding to the multiple privacy identity ranges that are larger than the object identity range.

[0059] The replacement desensitization submodule 10246 is used to replace and desensitize the privacy identity information corresponding to multiple privacy identity ranges that are not larger than the object identity range.

[0060] Furthermore, the financial service platform based on privacy computing also includes: The model training and verification unit 103 is used to select a basic privacy computing model, collect data through multiple targets, perform model training and verification on the basic privacy computing model, and generate a target privacy computing model.

[0061] In an embodiment of the present invention, the model training and verification unit 103 selects a basic privacy computing model according to the financial service business. The basic privacy computing model specifically includes: a multi-party secure computing model, a homomorphic encryption model or a differential privacy model, etc., and then obtains the data division ratio, generates a data division instruction according to the data division ratio, and then divides and processes the data collected from multiple targets according to the data division instruction to obtain a test set and a verification set. The basic privacy computing model is trained through the test set, and then the basic privacy computing model is verified through the verification set to generate a target privacy computing model.

[0062] Specifically, Figure 6 A structural block diagram of the model training verification unit 103 in the system provided by an embodiment of the present invention is shown.

[0063] Among them, in the preferred embodiment provided by the present invention, the model training verification unit 103 specifically includes: The basic model selection module 1031 is used to select a basic privacy computing model according to the financial service business.

[0064] The collected data division module 1032 is used to divide the plurality of target collected data into a test set and a validation set.

[0065] Specifically, Figure 7 It shows a structural block diagram of the data collection and partitioning module 1032 in the system provided by an embodiment of the present invention.

[0066] Among them, in the preferred embodiment provided by the present invention, the collected data division module 1032 specifically includes: The ratio acquisition submodule 10321 is used to acquire the data division ratio.

[0067] The instruction generation submodule 10322 is used to generate a data partitioning instruction according to the data partitioning ratio.

[0068] The partition processing submodule 10323 is used to partition the plurality of target collected data according to the data partition instruction to obtain a test set and a verification set.

[0069] Furthermore, the model training verification unit 103 also includes: The model training and verification module 1033 is used to perform model training and verification on the basic privacy computing model through the test set and the verification set to generate a target privacy computing model; The loss function used in the model training process optimizes the model parameters. The loss function expression is: ; in, represents the optimal value of the model parameters, represents the model parameter vector, represents the number of samples of training data, Indicates loss, Represents the prediction result of the model, represents the data input of a single sample, represents the true label vector, , and Both represent regularization parameters, represents the L2 norm of the parameter, represents the gradient of the loss function with respect to the model parameters, represents the dimension of the model parameters, represents the sum of squares of the second-order derivatives of the loss function, represents the symbol of partial derivative, represents the first parameters.

[0070] Specifically, L2 regularization is introduced to prevent model overfitting and ensure that the parameter amplitude is moderate. At the same time, gradient changes are controlled to avoid excessive parameter adjustments during training and improve the stability of the model.

[0071] Furthermore, the financial service platform based on privacy computing also includes: The model deployment application unit 104 is used to deploy the target privacy computing model in the financial service business, and perform application monitoring and adjustment optimization on the target privacy computing model.

[0072] In an embodiment of the present invention, in the financial services business, the target privacy computing model is deployed through the model deployment application unit 104 to realize the application of the target privacy computing model in the financial services business. During the application process, the target privacy computing model is monitored and application monitoring data including calculation accuracy, response time and response exceptions are recorded. Then, based on the application monitoring data, the target privacy computing model is adjusted and optimized to improve the efficiency and stability of the target privacy computing model.

[0073] Specifically, Figure 8 It shows a structural block diagram of the model deployment application unit 104 in the system provided by an embodiment of the present invention.

[0074] Among them, in the preferred implementation manner provided by the present invention, the model deployment application unit 104 specifically includes: The model deployment module 1041 is used to deploy the target privacy computing model in the financial services business.

[0075] The application monitoring module 1042 is used to perform application monitoring on the target privacy computing model and record application monitoring data including computing accuracy, response time and response anomalies.

[0076] An adjustment and optimization module 1043, configured to adjust and optimize the target privacy computing model according to the application monitoring data; Among them, the relationship existing in the process of adjusting and optimizing the target privacy computing model is: ; in, represents the change in model parameters, represents the learning rate, represents the second-order derivative matrix of the loss, represents the true label, represents the gradient of loss with respect to model parameters, represents the weight of the momentum term, represents the momentum term, represents a constant used to adjust the optimization results, and Both represent coefficients used to adjust the exponential function. Represents input data, Represents the gradient operator with respect to a parameter.

[0077] Specifically, the second-order derivative matrix of the loss is used to provide the second-order derivative information of the loss function to accelerate the convergence speed and improve the optimization effect; at the same time, the momentum term is introduced to help overcome the local minimum and improve the convergence speed.

[0078] Furthermore, the financial service platform based on privacy computing also includes: The model iteration and updating unit 105 is used to continuously collect multiple relevant process data in the financial service business, iterate the target privacy computing model through the multiple relevant process data, and update the iterative privacy computing model.

[0079] In an embodiment of the present invention, in the financial services business, the model iteration update unit 105 continuously collects multiple relevant process data, analyzes the multiple relevant process data, screens multiple valid process data from the multiple relevant process data, and then iterates the target privacy computing model through multiple valid process data and / or introduces new privacy computing technology, updates the iterative privacy computing model, thereby improving the performance and security of financial services based on privacy computing.

[0080] Specifically, Fig. 9 It shows a structural block diagram of the model iteration updating unit 105 in the system provided by an embodiment of the present invention.

[0081] Among them, in the preferred implementation manner provided by the present invention, the model iterative updating unit 105 specifically includes: The continuous collection module 1051 is used to continuously collect a plurality of relevant process data in the financial service business.

[0082] The data screening module 1052 is used to screen a plurality of valid process data from the plurality of related process data.

[0083] A model iteration module 1053 is used to iterate the target privacy computing model through a plurality of valid process data, and update the iterative privacy computing model; Among them, the relationship in the process of iterating the target privacy computing model is: ; in, Indicates The model parameters of the round iteration, Indicates The model parameters of the round iteration, represents the learning rate, Indicates The gradient of the round iteration, represents the regularization parameter, represents the regularization term, Represents a constant used to adjust the iteration result. Represents the coefficient used to adjust the logarithmic function.

[0084] Specifically, by introducing gradient descent, regularization and momentum method, the model complexity and learning rate can be effectively controlled, and the convergence and model performance can be improved; at the same time, through the regularization term, different types of regularization can be selected to adapt to different model requirements.

[0085] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0086] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0087] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0089] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A financial service platform based on privacy computing, characterized in that: The platform includes a service request processing unit, a data collection processing unit, a model training and verification unit, a model deployment and application unit, and a model iteration and update unit, wherein: A service request processing unit, configured to receive a financial service request, analyze the financial service request, determine a service request object and a financial service business, and generate and sign a financial service agreement with the service request object; A data collection and processing unit, configured to collect data according to the financial service agreement, obtain a plurality of relevant collected data, and perform standardization and range desensitization processing on the plurality of relevant collected data to generate a plurality of target collected data; A model training and verification unit, used to select a basic privacy computing model, collect data through multiple targets, perform model training and verification on the basic privacy computing model, and generate a target privacy computing model; A model deployment application unit, used to deploy the target privacy computing model in the financial service business, and to perform application monitoring and adjustment optimization on the target privacy computing model; The model iteration and updating unit is used to continuously collect multiple related process data in the financial service business, iterate the target privacy computing model through the multiple related process data, and update the iterative privacy computing model.

2. The financial service platform based on privacy computing according to claim 1, characterized in that: The service request processing unit specifically includes: A service request receiving module, used for receiving a financial service request; A service request analysis module, used to analyze the financial service request and determine the service request object and the financial service business; A service agreement generation module, used to generate a financial service agreement; A service agreement signing module is used to sign the financial service agreement with the service request object.

3. The financial service platform based on privacy computing according to claim 2, characterized in that: The service agreement generation module specifically includes: A service analysis submodule, for clarifying service responsibilities, service rights and service scope according to the financial service business; A service terms formulation submodule is used to formulate relevant service terms according to the service responsibilities, service rights and service scope; The legal and regulatory matching submodule is used to match relevant laws and regulations; The service agreement creation submodule is used to integrate the relevant service terms and the relevant laws and regulations to create a financial service agreement.

4. The financial service platform based on privacy computing according to claim 3, characterized in that: The data collection and processing unit specifically includes: A related type determination module, used to determine a plurality of related data types according to the financial service agreement; A related data collection module, used to collect data according to a plurality of related data types, and obtain a plurality of related collected data; A standardization processing module, used for performing standardization processing on the plurality of related collected data to obtain a plurality of standard collected data; Among them, the relationship between the standardization of the plurality of related collected data is: ； in, represents the original data matrix, represents the mean vector of the data, represents the covariance matrix of the data, represents the standardized data matrix, and Both represent constants used to adjust the standardized results. represents a constant used to avoid zero denominator in logarithmic operations. represents element-wise multiplication; A range desensitization processing module, used for performing range desensitization processing on a plurality of the standard collection data to generate a plurality of target collection data; Among them, the relationship formula that exists in the process of performing range desensitization processing on multiple standard collection data using the range desensitization processing module is: ； in, Represents the desensitized data matrix, represents the mask matrix, represents the result of hashing the data matrix, represents the Hadamard product, Represents a constant used to adjust the hashing result. and Both represent the coefficients used to adjust the cosine function.

5. The financial service platform based on privacy computing according to claim 4, characterized in that: The range desensitization processing module specifically includes: An object identity range determination submodule, used to determine the object identity range of the service request object; A privacy identification submodule, used to perform privacy identification on a plurality of the standard collected data and mark a plurality of privacy identity information; A privacy identity range identification submodule, used to identify the privacy identity ranges of the plurality of privacy identity information; A range comparison submodule, configured to compare a plurality of the privacy identity ranges with the object identity range; A retention processing submodule, used for retaining the privacy identity information corresponding to the multiple privacy identity ranges larger than the object identity range; The replacement desensitization submodule is used to replace and desensitize the privacy identity information corresponding to multiple privacy identity ranges that are not larger than the object identity range.

6. The financial service platform based on privacy computing according to claim 5, characterized in that: The model training verification unit specifically includes: A basic model selection module, used to select a basic privacy computing model according to the financial service business; A collection data division module, used for dividing the plurality of target collection data into a test set and a validation set; A model training and verification module, used to perform model training and verification on the basic privacy computing model through the test set and the verification set to generate a target privacy computing model; The loss function used in the model training process optimizes the model parameters. The loss function expression is: ； in, represents the optimal value of the model parameters, represents the model parameter vector, represents the number of samples of training data, Indicates loss, Represents the prediction result of the model, represents the data input of a single sample, represents the true label vector, , and Both represent regularization parameters, represents the L2 norm of the parameter, represents the gradient of the loss function with respect to the model parameters, represents the dimension of the model parameters, represents the sum of squares of the second-order derivatives of the loss function, represents the symbol of partial derivative, represents the first parameters.

7. The financial service platform based on privacy computing according to claim 6, characterized in that: The data collection and partitioning module specifically includes: The ratio acquisition submodule is used to obtain the data division ratio; An instruction generation submodule, used for generating a data partitioning instruction according to the data partitioning ratio; The partitioning processing submodule is used to partition the data collected from the plurality of targets according to the data partitioning instruction to obtain a test set and a verification set.

8. The financial service platform based on privacy computing according to claim 7, characterized in that: The basic privacy computing model specifically includes: a multi-party secure computing model, a homomorphic encryption model or a differential privacy model.

9. The financial service platform based on privacy computing according to claim 8, characterized in that: The model deployment application unit specifically includes: A model deployment module, used to deploy the target privacy computing model in the financial service business; An application monitoring module, used to perform application monitoring on the target privacy computing model and record application monitoring data including computing accuracy, response time and response anomalies; An adjustment and optimization module, used to adjust and optimize the target privacy computing model according to the application monitoring data; Among them, the relationship existing in the process of adjusting and optimizing the target privacy computing model is: ； in, represents the change in model parameters, represents the learning rate, represents the second-order derivative matrix of the loss, represents the true label, represents the gradient of loss with respect to model parameters, represents the weight of the momentum term, represents the momentum term, represents a constant used to adjust the optimization results, and Both represent the coefficients used to adjust the exponential function. Represents input data, Represents the gradient operator with respect to a parameter.

10. The financial service platform based on privacy computing according to claim 9, characterized in that: The model iteration updating unit specifically includes: A continuous collection module, used for continuously collecting a plurality of relevant process data in the financial service business; A data screening module, used for screening a plurality of valid process data from a plurality of the related process data; A model iteration module, used to iterate the target privacy computing model through a plurality of valid process data, and update the iterative privacy computing model; Among them, the relationship in the process of iterating the target privacy computing model is: ； in, Indicates The model parameters of the round iteration, Indicates The model parameters of the round iteration, Indicates The gradient of the round iteration, represents the regularization parameter, represents the regularization term, represents a constant used to adjust the iteration result, Represents the coefficient used to adjust the logarithmic function.

Images