Privacy protection method and device, electronic equipment and storage medium

Abstract

This application relates to a privacy protection method, apparatus, electronic device, and storage medium, and pertains to the fields of data security and financial technology. The method includes: acquiring raw business data and corresponding business tools from a business scenario; decomposing the business processing flow of the raw business data and business tools using preset element-based processing routing rules to obtain business data elements corresponding to the raw business data and tool processing elements corresponding to the business tools; responding to a business operation instruction corresponding to the business scenario, determining the target operation element corresponding to the business operation instruction from the business data elements and tool processing elements; determining the privacy operation corresponding to the target operation element based on its element category and scheduling flow; encrypting the target operation element using the privacy operation to assemble the encrypted target processing flow; and generating the business processing result corresponding to the business operation instruction.

Description

Technical Field

[0001] This application relates to the fields of data security and financial technology, specifically to a privacy protection method, device, electronic device, and storage medium. Background Technology

[0002] In financial business scenarios, such as valuation and guarantee businesses, privacy computing technology can often be used to protect business data, solve data security issues, and build a secure and feasible business operating environment.

[0003] Related technologies typically employ methods such as multi-party secure computation, federated learning, and trusted execution environments to encrypt and isolate business data and mask its privacy.

[0004] However, in order to achieve specific business functions, financial institutions usually use business tools to configure custom business rules to further process business data. The relevant technologies can usually only protect the business data, but cannot protect the custom business rules of the business data. This leads to the risk of leakage in the calculation process of business data and cannot guarantee the overall security of the business processing flow. Summary of the Invention

[0005] In view of this, this application provides a privacy protection method, device, electronic device and storage medium, the main purpose of which is to solve the technical problem that related technologies can usually only protect business data, but cannot protect the customized business rules of business data, resulting in the risk of leakage in the calculation process of business data and failing to guarantee the overall security of the business processing flow.

[0006] According to a first aspect of this application, a privacy protection method is provided, the method comprising: Acquire raw business data and corresponding business tools from the business scenario; By pre-setting element-based processing routing rules, the business processing flow of the original business data and business tools is decomposed to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools. In response to the business operation instructions corresponding to the business scenario, the target operation element corresponding to the business operation instruction is determined from the business data elements and tool processing elements; Based on the element category and scheduling process of the target operation element, determine the privacy operation corresponding to the target operation element, use the privacy operation operation to encrypt the target operation element, and assemble the encrypted target processing flow. Based on the encrypted target processing flow, the business processing results corresponding to the business operation instructions are generated.

[0007] According to a second aspect of this application, a privacy protection device is provided, the device comprising: The acquisition module is used to acquire raw business data and the corresponding business tools in the business scenario. The routing module is used to decompose the business processing flow of the original business data and business tools through preset element-based processing routing rules, so as to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools. The determination module is used to determine the target operation element corresponding to the business operation instruction from the business data elements and tool processing elements in response to the business operation instruction corresponding to the business scenario. The encryption module is used to determine the privacy operation corresponding to the target operation element based on the element category and scheduling process of the target operation element, and to encrypt the target operation element using the privacy operation operation to assemble the encrypted target processing flow. The processing module is used to generate business processing results corresponding to business operation instructions based on the encrypted target processing flow.

[0008] According to a third aspect of this application, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method of the first aspect described above.

[0009] According to a fourth aspect of this application, a computer-readable storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements the method of the first aspect described above.

[0010] Compared with related technologies, the privacy protection method, apparatus, electronic device, and storage medium provided in this application can obtain the original business data and the corresponding business tools in a business scenario; decompose the business processing flow of the original business data and business tools by pre-set element-based processing routing rules to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools; respond to the business operation instructions corresponding to the business scenario, determine the target operation element corresponding to the business operation instruction from the business data elements and tool processing elements; determine the privacy operation corresponding to the target operation element according to the element category and scheduling flow of the target operation element, encrypt the target operation element using the privacy operation, and assemble the encrypted target processing flow; and generate the business processing result corresponding to the business operation instruction based on the encrypted target processing flow. In this way, upon receiving a business operation instruction, the required target operation elements are retrieved from business data elements and tool processing elements. The privacy-preserving computational operations corresponding to each operation element are then encrypted. The encrypted target operation elements are used to assemble a fully encrypted target processing flow. The business operation instruction is processed according to this encrypted flow to generate the corresponding business processing result. This eliminates the need to directly use the original business data and business processing flow, achieving element-based dual control over the original business data and business tools. This, in turn, provides dual privacy protection for the original business data and business processing flow, reduces the risk of leakage of original business data and business rules, and meets the requirements for efficient business processing and strong privacy protection. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A schematic flowchart illustrating a privacy protection method, apparatus, electronic device, and storage medium provided in an embodiment of this application; Figure 2 This is a schematic diagram of the overall architecture of an example provided in an embodiment of this application; Figure 3 A schematic diagram illustrating the operation mode of an example element-based routing device provided in an embodiment of this application; Figure 4A schematic diagram illustrating a client-side multi-stage privacy protection method, provided as an example of an embodiment of this application; Figure 5 A schematic diagram of an example data anonymity query architecture provided in an embodiment of this application; Figure 6 A schematic diagram of an example fully dense-state computation architecture provided in an embodiment of this application; Figure 7 This is a schematic diagram of a privacy protection device provided in an embodiment of this application. Detailed Implementation

[0014] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0015] The privacy protection methods, apparatus, electronic devices, and storage media of embodiments of this application are described below with reference to the accompanying drawings.

[0016] In financial operations, financial institutions typically need to process massive amounts of financial business data and utilize financial engineering tools for business processing, such as innovating financial products, formulating quantitative investment strategies, strengthening risk management, forecasting trends, and improving the efficiency of risk measurement and the accuracy of product pricing.

[0017] Related technologies typically employ homomorphic encryption, differential privacy, secret sharing, multi-party secure computation, and trusted execution environments to protect financial data. However, differential privacy offers relatively low protection levels, and homomorphic encryption suffers from low computational efficiency. Different protection methods exhibit varying performance in terms of privacy protection levels, ciphertext computation efficiency, and accuracy loss, with no universal solution. Furthermore, privacy protection remains limited to business data. Business data serves as the foundation for decision-making, while financial engineering tools (financial engineering models) act as the econometric engine, together forming the core system of intelligent decision-making and risk management for financial institutions. Existing privacy-preserving computation technologies focus solely on protecting the privacy of business data, neglecting to protect these core econometric engines like financial engineering tools. Consequently, these technologies fail to protect the privacy of the computational process within financial engineering tools, cannot achieve dual privacy protection for both data and models, and cannot strike a balance between the quality of privacy protection and the timeliness of computational services.

[0018] This application provides a privacy protection method, device, electronic device, and storage medium. Its main purpose is to solve the technical problem that related technologies can usually only protect business data, but cannot protect the customized business rules of business data, resulting in the risk of leakage in the calculation process of business data and failing to guarantee the overall security of the business processing flow.

[0019] like Figure 1 As shown, embodiments of this application provide a privacy protection method, apparatus, electronic device, and storage medium, including: Step 101: Obtain the original business data and the corresponding business tools in the business scenario.

[0020] The business scenarios can include financial business scenarios, such as credit, accounting, trade services, and payments. Raw business data can include directly obtained business data that requires privacy protection, such as: business process data and behavioral data generated by customers using financial institution services; structured data stored in relational databases; unstructured data stored in file systems, such as credit reports, risk ratings, digital certificates, and customer service dual recordings; and market data and pricing time-series data provided by other financial intermediaries and information service providers, stored in in-memory databases.

[0021] Correspondingly, business tools can be business processing models used by business organizations in the current business scenario, such as pricing models, measurement models, risk control models, etc. These models can include multiple calculation steps determined based on business data and custom business rules.

[0022] Step 102: Decompose the business processing flow of the original business data and business tools by using preset element-based processing routing rules to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools.

[0023] In this embodiment, business data elements can be the smallest data units obtained by decomposing business data, and tool processing elements can be the smallest functional units obtained by decomposing business processing flows. Business data elements and tool processing elements can be formal business elements after standardized transformation and governance. They can process business operation requests after scheduling and combination. For example, using encryption algorithms, based on element definition information and scheduling information, business data items and tool processing items can be converted into business data elements and tool processing elements, and then stored in categories to protect the privacy of the original business data items and tool processing items. The execution entity in this embodiment can be a server, which can use a database to classify and store business data elements and tool processing elements.

[0024] In some embodiments, the preset element-based processing routing rules can be pre-configured standardized routing rules for element processing to achieve automated routing of raw business data and business processing flows. Business processing flows may include the calculation process of business tools, data flow processes, etc. For example, atomic data splitting can be performed based on the data source and data type of the raw business data to obtain business data items to be processed, such as position quantity, transaction price, and transaction time; and atomic steps can be split according to the specific calculation steps, calculation formulas, and model operation processes in the business processing flow to divide the continuous and complete business chain into multiple tool processing items, where each tool processing item can be a single atomic calculation action.

[0025] Step 103: In response to the business operation instructions corresponding to the business scenario, determine the target operation element corresponding to the business operation instruction from the business data elements and tool processing elements.

[0026] The target data elements may include at least one business data element that needs to participate in this operation, which is selected and matched according to the requirements of the current business operation instructions, and at least one tool processing element required for this process, which is selected and matched.

[0027] For example, when a business operation instruction is received from the client, the business requirements of the business operation instruction are parsed, and the target operation elements required for this operation are determined from business data elements and tool processing elements through keyword matching.

[0028] Step 104: Based on the element category and scheduling process of the target operation element, determine the privacy operation corresponding to the target operation element, use the privacy operation operation to encrypt the target operation element, and assemble the encrypted target processing flow.

[0029] The target processing flow can be the processing flow involved in the current business operation instruction in the business processing flow of the business tool.

[0030] In some embodiments, the element category and scheduling process of each target operation element can be identified, the original data type and original processing flow of the original business data corresponding to each target operation element can be determined, the privacy operation of each target operation element, i.e., the encryption algorithm, can be selected according to the original data type and original processing flow, and the target operation element can be encrypted to obtain the encrypted target operation element. Then, it can be assembled according to the original processing flow corresponding to the target operation element to generate the encrypted target processing flow, thereby reducing the risk of leakage of the specific processing logic rules of the original business data and business tools.

[0031] Step 105: Based on the encrypted target processing flow, generate the business processing result corresponding to the business operation instruction.

[0032] In some embodiments, the encrypted target processing flow is consistent with its corresponding original processing flow, except that the data used is encrypted. Finally, the target processing flow can be executed to generate the business processing result corresponding to the business operation instruction.

[0033] Compared with related technologies, this embodiment can extract business data elements from the original business data of the business scenario and decompose the business processing flow of the business tool to obtain tool processing elements. When a business operation instruction is received, the required target operation elements are called from the business data elements and tool processing elements, and the privacy operation corresponding to each operation element is determined and encrypted. The encrypted target operation elements are used to assemble a fully encrypted target processing flow. The business operation instruction is processed according to the encrypted flow to generate the business processing result corresponding to the business operation instruction. It does not require the direct use of the original business data and business processing flow, and realizes element-based dual control of the original business data and business tools. This achieves dual privacy protection for the original business data and business processing flow, reduces the risk of leakage of the original business data and business rules, and meets the requirements of efficient business processing and strong privacy protection.

[0034] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, and to fully illustrate the implementation of this embodiment, step 102 may optionally include: configuring preset element-based processing routing rules based on the data characteristics of the original business data and the tool processing characteristics of the business processing flow; matching business keywords corresponding to the original business data and the business processing flow based on the preset element-based processing routing rules, and determining the routing path corresponding to the original business data and the business processing flow based on the keyword matching results; and decomposing the original business data and the business processing flow according to the routing path corresponding to the original business data and the business processing flow to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools.

[0035] For example, such as Figure 2As shown, taking the financial business scenario corresponding to the assets of financial institutions as an example, business data can be routed to the business data elementization module through the elementization routing module (elementization processing routing module), and intermediate data corresponding to financial engineering (financial engineering) tools can be routed to the tool elementization module (such as the financial engineering tool elementization module). The business data elementization module can be divided into a raw data layer, a basic data layer, a data mining layer, and a data governance layer based on the data governance process. The raw data layer can be used to transmit self-collected data sources, purchased data sources, and internal proprietary data sources using a bidirectional long-connection communication protocol (WebSocket), a distributed real-time stream computing engine (Flink), and a binary change log (Binlog), respectively. Data from multiple sources is aggregated through a message middleware (distributed message queue Kafka), and the collected raw business data is transmitted to the basic data layer. The basic data layer can be used for data cleaning and classification storage of the raw business data. The system can be categorized into multiple basic databases, such as a basic database, a business database, a log database, and a monitoring database. The data mining layer can extract data based on these categorized basic databases. The big data platform handles task scheduling and ETL (Extract, Transform, Load) task computation. It can build analysis databases, graph databases, indicator databases, and tag databases corresponding to business data elements based on the data source layer (ODS), common model layer (CMD), and application layer (ADS). The data governance layer provides data governance-related functions, including key governance tasks such as data quality, data monitoring, data standards, data transmission, and data collection, covering the entire governance service capability from data collection to final element generation. The governance service, based on data transmission capabilities, enables interaction and connection with components of the raw data layer, basic data layer, and data mining layer, continuously improving the real-time performance of governance work.

[0036] Accordingly, the tool element module can be divided into an atomic tool layer and an operation scheduling layer based on the tool governance process. Data governance can include processes such as operational quality, tool monitoring, splitting standards, tool analysis, and tool data collection. Quantitative tools can include customized models (such as quantitative investment strategy models, risk measurement management models, and financial product pricing models), basic models (such as self-developed models and Monte Carlo models), and basic formulas (such as calculation formulas for indicators like Value at Risk, Sharpe ratio, maximum drawdown, market volatility, and duration). The atomic tool layer can be used for extracting and splitting element information of quantitative tools. The split atomic data can be stored in a decomposition library according to processing categories. The decomposition library can include an atomic formula library, a calculation parameter library, and a calculation step library. The operation scheduling layer can be used for formula lineage graph construction, calculation data acquisition, atomic tool acquisition, and scenario-scheduled calculations.

[0037] In some embodiments, the element-based routing module can be used to configure preset element-based processing routing rules, gating routing rules, intent recognition rules, and reinforcement learning rules to provide a combined adaptive routing strategy. The original business data is then routed to the business data element-based module, and the business processing flow is routed to the tool element-based module. Specifically, the preset element-based processing routing rules can be configured based on business data characteristics and tool characteristics, such as keywords, semantic categories, and element complexity. This rule method function can be represented as Router.Rule(), used to construct the element-based processing steps.

[0038] Further optionally, based on preset element-based processing routing rules, the system matches business keywords corresponding to the original business data and business processing flow, and determines the routing path corresponding to the original business data and business processing flow based on the keyword matching results. Specifically, this may include: identifying business keywords corresponding to the original business data and the process data of the business processing flow; using rule-based routing based on the matching degree between the preset element-based processing routing rules and the business keywords, as well as the complexity of the business keywords, to generate rule-based routing processing steps corresponding to the original business data and business processing flow; using gated routing to convert the original business data and process data into embedded vectors, and based on the compatibility matching degree between the embedded vectors and the preset gated processing paths in the gated routing, determining the routing path based on compatibility matching... The system determines at least one processing path corresponding to the original business data and process data, respectively, and performs weighted aggregation on the at least one processing path to determine the gated routing processing steps corresponding to the original business data and process data, respectively. Using an intent recognition model, it identifies the expected execution targets corresponding to the original business data and process data, and generates intent recognition processing steps based on the expected execution targets. Using a reinforcement learning model, it performs rule reinforcement learning based on the rule-based routing processing steps, gated routing processing steps, and intent recognition processing steps to generate reinforcement learning processing steps. Based on at least one of the rule-based routing processing steps, gated routing steps, intent recognition processing steps, and reinforcement learning processing steps, it determines the routing paths corresponding to the original business data and process data, respectively.

[0039] For example, such as Figure 3As shown, the system first allows input of raw business data and processing flows from the business scenario. Using a keyword matching method (Router.Rule.KeywordMatch()), semantic classification is performed based on natural language processing technology (Router.Rule.NlpProcess()). Complexity is assessed based on the number of elements and term density (Router.Rule.DifficlutyCal()), resulting in corresponding business process handling steps (OperationProcessRuleList). For example: Entering the business data element module -> entering the raw data layer -> clarifying data governance rules -> entering the basic data layer via data transmission synchronization -> performing data mining; entering the business data element module -> entering the basic data layer -> clarifying the financial tool governance rules -> entering the data mining layer; entering the tool element module -> forming atomic formulas -> clarifying data governance rules -> performing the operation scheduling layer, forming different routing paths through rule routing to determine the handling methods and steps for different data. For gated routing (gated networks), the corresponding method function can be represented as Router.Gate(), which can convert input data into embeddings. Based on the converted embeddings, it calculates the compatibility score (CalScore) between the input data and all element-based processing paths, reflecting the compatibility matching degree between the input elements and the processing paths (OperationProcess). Based on the CalScore, it selects the top-K processing paths with the highest scores for each input data, and then weights and aggregates the K paths to form the corresponding element-based processing steps, such as the business process gateway processing steps (OperationProcessGateList). For the intent recognition model, the corresponding method function can be represented as Router.Recognize(), which can be a large language model. Based on the semantic understanding capability of the large language model, prompt words can be written to guide the model to analyze the classification identifiers or instructions of the input and output element processing path. According to the expected execution goal of the next step, the execution workflow corresponding to the expected execution goal identified by the model is directed to the corresponding processing path, forming the business process recognition and handling steps (OperationProcessRecognizeList) corresponding to intent recognition. For the reinforcement learning module, the corresponding method function can be represented as Router.RL(), which can construct a routing agent. Through the routing agent, reinforcement learning is performed on the routing processes of different elements in different business scenarios. For example, it can obtain business execution environment information in the current business scenario, such as system resources and the name of the business processing model, and obtain multiple element-based processing path results of the element-based routing module, such as OperationProcessRuleList, OperationProcessGateList, and OperationProcessRecogeList. Based on the business execution environment information and multiple element-based processing path results, the preset element-based processing routing rules are continuously optimized, the routing strategy is gradually optimized, and the optimized reinforcement learning processing steps (OperationProcessRLList) are generated. Finally, the element-based routing module outputs element-based processing steps (OperationProcessList) for filtering to determine the final routing paths corresponding to the original business data and process data respectively. For example, if both OperationProcessRuleList and OperationProcessGateList appear, then OperationProcessRuleList will be used as the final output; if OperationProcessRLList also appears, then OperationProcessRLList will be used as the final output.

[0040] In some embodiments, a multi-path enhancement mechanism based on rules, gating networks, and intent recognition can be added to Q-Learning or policy gradient algorithms. That is, in addition to the algorithm itself, the operation results of the other three routing mechanisms are also used to form path information to enhance the usability of the algorithm, thereby enabling the automatic selection of model combinations based on business operation instructions through the routing agent.

[0041] Optionally, acquiring raw business data from the business scenario may include: acquiring raw business data from different sources within the business scenario through the raw data layer; identifying the encryption state corresponding to the raw business data, and determining the target transmission channel corresponding to the raw business data based on the encryption state; using the target component corresponding to the target transmission channel, transmitting the raw business data to the basic data layer through a message middleware; and cleaning and classifying the raw business data through the basic data layer to obtain the classified and stored raw business data. Here, the target transmission channel can be a transmission channel corresponding to data from different sources, and may include multiple preset channels; the target component can be a preset component for processing data from different sources, such as a Flink component, a Binlog component, etc.; the data source may include self-collected data sources, purchased data sources, internal proprietary data sources, etc.; and the message middleware may include a Kafka device.

[0042] Correspondingly, the business data element module can be used to encapsulate data element components using standard processes, specifically including steps such as data aggregation processing, element component processing, and element component governance. Data aggregation processing can be performed through the raw data layer and the basic data layer. The raw data layer can simultaneously collect raw business data from different sources, while the basic data layer can perform cleaning, quality inspection, classification, and grading of the raw business data.

[0043] At the raw data layer, different communication protocols can be used for data transmission due to the different sources of raw business data. For example, for self-collected data sources, a TWS protocol specifically for real-time transmission of task information can be built based on WebSocket technology. This involves expanding the existing WebSocket protocol with data source tags (SourceTag) and task scheduling (TaskSchedule) information, reconstructing the original WebSocket protocol methods, and inheriting the channel opening callback method (onopen()) to form the task channel opening method (onTaskopen()), which opens the task data transmission channel using task graph information (TaskGraph) as input. It also inherits the message receiving callback method (onmessage()) to form the task message processing method (onTaskmessage()), which performs data scheduling based on the TaskGraph and calls different business data transmission channels according to the data type. For structured or streaming data, the Flink component is called based on the Flink task invocation method (taskFlink()), and for log or file data, the corresponding self-developed Binlog component is called based on the Binlog task invocation method (taskBinlog()) for data transmission.

[0044] For procurement data sources, Flink components can be built to provide high-quality transmission services for both structured and streaming data. Product expansion based on open-source Flink components includes: first, adding a collaboration interface (taskInvoke()) with the WebSocket task scheduling component, supporting real-time scheduling and guidance of Flink components through the task scheduling component, and real-time synchronization of task information with the task scheduling device through the task call result synchronization interface (taskInvokeRes()) after task completion; second, optimizing the existing Flink partitioning design and adding extended support for SourceTag to differentiate data source channels.

[0045] For internal data sources, a Binlog component can be built to provide high-quality transmission services for unstructured data such as file logs. Based on the open-source Flink Change Data Capture component (FlinkCDC), and drawing heavily on the design pattern of MySQLBinlog, the component design is completed. First, a collaborative interface, `taskInvoke()`, is added for the WebSocket task scheduling component, supporting real-time scheduling and guidance of the Binlog component through the task scheduling component. Second, the Binlog operation mode is constructed. First, the structure information extraction method (StrucExtract()) is used to parse the log or file structure to obtain structure information (StructureInfo). Then, the unstructured data parsing and transmission task is started by marking the starting scheduling point using the transfer scheduling method (TransferScheduler()). The parsed unstructured data first enters the binlog cache and is marked with the corresponding SourceTag. After the Binlog synchronization method (syn_binlog()) converts the data in the cache into result information according to the StructureInfo, it is transmitted to the destination. Finally, after the task is completed, the task information is synchronized in real-time with the task scheduling device through the `taskInvokeRes()` interface.

[0046] Optionally, the raw data layer can also be configured with a Kafka device to expand existing partitions, add task scheduling type (SchedulerType) and data source label, realize the integration of business data with different scheduling modes, and input the integrated data from multiple sources into the basic data layer.

[0047] For the basic data layer, the data type and data characteristics of the original business data can be identified. Based on the data type and data characteristics, the corresponding cleaning method for the original business data can be determined. The original business data is then cleaned according to the cleaning method to obtain the cleaned business data, which is then converted into structured storage data. The original business data may include business data from different sources, such as self-collected, externally purchased, and internally owned data. For business data from different sources, source identification can be performed before data transmission.

[0048] For example, for structured and streaming business data, as well as unstructured file data, data preprocessing is performed by constructing feature engineering methods tailored to different data types. Through three steps—feature selection, feature extraction, and feature construction—data features of the original business data can be extracted and constructed based on the data type and business category of the original business data. For example, based on the features, missing value handling (filling in the values ​​of the features belonging to the category), outlier and duplicate value removal (identification based on feature information), data consistency processing, and data normalization / standardization (operations based on the feature distribution of the features) can be performed sequentially. This embodiment first forms feature data through feature engineering, and then performs data cleaning based on the data features. This ensures that the cleaned data, whether originally structured or unstructured, is transformed into structured storage data, giving unstructured information traceability capabilities. For example, a source link can be added to the structured information to associate the original unstructured information with the structured information. Subsequently, data is stored separately according to business categories such as basic data, business data, log data, and monitoring data.

[0049] Through a multi-layered business data element module, a multi-mode collaborative data transmission synchronization solution can be provided. It uses different methods such as TWS protocol, upgraded Flink, and self-developed Binlog to transmit business data from different sources. It also uses aspect-oriented programming (AOP) aspect design to realize the synchronous execution of element definition calculation and data transmission, build atomic capabilities for business data elements, and achieve privacy protection at the atomic granularity of business data.

[0050] Optionally, the original business data and business processing flow can be decomposed to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools. Specifically, this may include: using a data extraction tool configured with preset entry points, during the transmission of the original business data, extracting business data items in batches based on element definition tool classes and scheduling task tool classes, and performing element definition and task scheduling processing on the business data items to generate business data elements after element-based transformation of the business data items. The preset entry points can be at least one preset entry point based on AOP for data from different sources.

[0051] The data mining layer can be used for element processing, and may include stages such as data element development, data element pilot testing, and data element production, corresponding to steps such as the data extraction tool DataX, task scheduling, and element definition in the data mining layer.

[0052] DataX comprises two key components: a data transmission module, consistent with the design of the data transmission module in the data acquisition layer; and an element AOP processing module, responsible for the implementation and application of element definition information and scheduling task data. Based on the AOP design principle, it constructs business data elements in real time during data transmission. First, in the DataX target class, corresponding entry points are designed, adding corresponding element definition operation functions, such as methods for adding, canceling, and deleting element definitions; and corresponding task scheduling operation functions, such as methods for adding, canceling, and deleting scheduling tasks. Then, corresponding element definition utility classes (DefinitionUtility) and scheduling task utility classes (SchdulerUtility) are constructed, defining transaction initiation and commit methods. Third, AOP weaving is completed by creating proxy objects to apply DefinitionUtility and SchdulerUtility to the target object DataX, achieving synchronous completion of data transmission and element generation.

[0053] Task scheduling can include two modes: task scheduling based on a conventional big data platform and custom ETL task computation. Corresponding data element components are designed and processed using both platform-native and custom methods, realizing a data processing flow from ODS to CMD to ADS. The presentation forms of element definitions can include analysis, graphs, indicators, and tags. By collecting corresponding business element information, different element definition forms are formed. During operation, element definition data and scheduling task information are transmitted to DataX together to complete data element governance.

[0054] In some embodiments, such as Figure 4As shown, the client can send business processing requests through the element interaction gateway to obtain business data elements, financial tool elements, and business execution processes. The received element process elements are then categorized and stored, such as business data, financial tools, and user data. The client can also perform element control, user management, and permission management. Element control can include resource access, usage registration, status management, and version management. The client can also perform encrypted element processing, such as configuring scene interaction functions for scene planning and result display; configuring computing services, such as indicator calculation and privacy computing; and configuring basic tools, such as element encryption tools and element decryption tools. Furthermore, it can perform end-to-end monitoring of business processing flows and requests within the business scenario, such as computation monitoring, storage monitoring, and process monitoring.

[0055] Optionally, data transmission adaptability can be improved by constructing a multi-mode collaborative data transmission synchronization scheme. If the encryption status of the original business data is identified as unencrypted, i.e., plaintext data, a solution combining TWS, self-developed Flink, and self-developed Binlog can be used to construct a real-time, bidirectional, customized data interaction channel between the client and the element gateway, as well as between element control, element storage, element processing, access control, and end-to-end monitoring within the client. Different interaction channels are generated according to different interaction content to achieve efficient information exchange. If the encryption state of the original business data is identified as encrypted, i.e., ciphertext data, a solution combining TWS, self-developed Flink, and self-developed Binlog can be used for ciphertext data transmission and interaction. Among them, the Secure Transmission Protocol (STWS) reconstructs the original TWS protocol method on the basis of TWS, and transmits all data after encryption with public key. Specifically, it inherits the onTaskopen method in TWS to form an encrypted task channel opening method (onSecretTaskopen()), which opens the task data transmission channel, using the task graph information as input; it inherits onmessage() to form an encrypted task message processing method (onSecretTaskmessage()), which performs data scheduling according to the TaskGraph, and calls different business data transmission channels according to different data categories. If it is structured or streaming data, it calls the Flink component in a ciphertext state based on the Flink encryption call method; if it is log or file data, it calls the corresponding self-developed Binlog component in a ciphertext state based on the Binlog encryption call method for data transmission.

[0056] In this way, the client can build corresponding module service capabilities according to the functional categories of element control, element storage, element processing, access control, and end-to-end monitoring. Furthermore, to build local encrypted computing capabilities on the client, sensitive data can be transmitted and used in computation via public-key encryption; and a hybrid privacy computing approach combining a software-implemented Trusted Execution Environment (STEE) and homomorphic encryption (HE) is proposed. Operators with high ciphertext computation efficiency, such as A+B, are directly computed using HE with homomorphic encryption; operators with low ciphertext computation efficiency, such as finding extrema, are computed based on STEE. STEE constructs an isolated Trusted Execution Environment implemented in a software model within the client terminal. Information and data in this environment are not shared with other modules on the client. All data entering the STEE is in plaintext, and all data leaving the STEE is in ciphertext. STEE only designs the kernel portion of the client as a minimum feasible computation base. This design is completed by the software itself and does not interact with the operating system or basic input / output system at the network or data layer, thus constructing an application virtual machine. Furthermore, on the client side, for highly complex data, a corresponding data virtual machine is built in STEE for storing the data in plaintext. For data of moderate complexity, public key encryption is used, and the encrypted data is stored in the client's encrypted data storage area.

[0057] In this way, a privacy computing service information transmission device can be built, using a combination of TWS, self-developed Flink, and self-developed Binlog to transmit plaintext and ciphertext data, and build multiple real-time data transmission channels.

[0058] As one possible implementation method, such as Figure 5 As shown, the upgraded secret sharing technology enables anonymous data querying, ensuring that data participates in computation in an efficient and secure manner. Taking data querying using sensitive information as an example, the specific process is as follows: (1) Server node B generates pseudo-random function (OPRF) seed s, calculates the corresponding secret sharing encrypted data for privacy data elements in the data elements of the server, such as sensitive information (such as bond identifiers), determines the OPRF calculation result, and uses the OPRF calculation result to replace the privacy data elements in the database; the database can be used to store basic data and provide computing services; data access and service calls can be performed between server node B and the database; (2) The client initiates an OPRF query request to the service node B based on the sensitive information it holds. For example, the client sends an OPRF query request to query m pieces of sensitive information. (3) The client and service node B interact with each other based on the OPRF protocol. Service node B generates a response to the OPRF query request and sends the OPRF values ​​corresponding to the sensitive information to the client. For example, it sends the OPRF values ​​corresponding to m pieces of sensitive information to the client, such as OPRF(s, ); (4) The client sends the OPRF values ​​corresponding to m sensitive information items and other data to the client agent node A; (5) Client agent node A and service node B can execute the RS-PSI protocol through the random secret sharing privacy set intersection (RS-PSI) module, so that the two nodes can obtain data fragments, such as secret sharing encrypted data of several query indicators, to realize secret sharing form of hidden query; (6) Client agent node A can return the plaintext data after decrypting the secret shared ciphertext data to the client.

[0059] As another possible implementation, multi-party secure computation capabilities based on secret sharing can be constructed to enhance the security of element privacy protection and achieve fully encrypted element computation, such as... Figure 6 As shown, the specific process is as follows: (1) Service node B generates OPRF seeds s, calculates OPRF for sensitive information (such as bond identifier ID) of all privacy data in financial institutions, and uses the obtained OPRF calculation results to replace the original sensitive information in the database; (2) The client initiates an OPRF query request based on the sensitive information it holds; (3) Service node B generates a response to the OPRF query request, namely the OPRF query result corresponding to the bond identifier ID required by the client (OPRF(s, )); (4) The client can send the OPRF corresponding to the bond identifier ID (key ID information) in the privacy data and other data to the client agent node A; (5) Client agent node A and service node B execute the RS-PSI protocol, and the execution result is a secret query RS-PSI module in the form of secret sharing. The two nodes obtain data shards. (6) The two nodes execute different operators of the secret sharing protocol, such as addition and multiplication, to perform calculations on the ciphertext and obtain the ciphertext result. (7) The data is decrypted by the client agent node A and returned to the client.

[0060] In this way, a combination of STEE-based and homomorphic encryption solutions can be provided to build client terminal privacy computing devices, provide data anonymity query technology services based on upgraded secret sharing, and achieve privacy protection through multi-party secure computing capabilities based on secret sharing.

[0061] In this embodiment, the element-based processing routing provides element category identification capabilities, transferring business data and business processing flows to the business data element-based module and the tool element-based module for unified standard element conversion, realizing the element-based conversion of business data and tools in business scenarios. Based on routing rules, gating networks, intent recognition, and reinforcement learning, a combined adaptive routing strategy is provided to improve the accuracy of element routing. Data elements are constructed based on data element-based governance, and a TWS protocol specifically for real-time transmission of task information is built based on WebSocket technology. Combined with an upgraded version of Flink and a self-developed version of Binlog, a data transmission module is constructed, resulting in two versions of the protocol: an encrypted version STWS and a plaintext version TWS. Data transmission and AOP aspect design are adopted to enable simultaneous data transmission and information computation. The routing scheme realizes the mapping between business processes and atomic capabilities, and implements client terminal privacy computation, upgraded data anonymity query, and expanded multi-party secure computation, achieving dual privacy protection for business data and business processing flows in business scenarios.

[0062] Optionally, the original business data and business processing flow are decomposed to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools. Specifically, this may also include: using data extraction tools, classifying the business processing flow according to the tool type during the execution of the business processing flow of the business tools, and obtaining the classified business process data; extracting tool processing items from the business process data in batches based on the element definition tool class and the scheduling task tool class, and performing element definition and task scheduling processing on the tool processing items to generate tool processing elements after the tool processing item element transformation.

[0063] For example, the tool element module can first collect financial engineering tool information according to categories such as basic formulas, basic models, and customized models. Through information extraction and decomposition at the atomic processing layer, atomic financial engineering tools such as atomic formula libraries, calculation parameter libraries, and calculation formula libraries are constructed. Finally, the formula lineage graph is constructed at the runtime scheduling layer, and functions such as calculation data acquisition, atomic formula acquisition, and scenario scheduling calculation are planned to support flexible scheduling of subsequent calculation processes. This forms the elements of financial engineering tools, which are then governed. DataX can be used for the transmission and processing of tool elements, such as tool element extraction, decomposition, and scheduling. The tool element governance service can complete the entire lifecycle of tool governance based on real-time transmission devices.

[0064] Correspondingly, the tool elementization module may include a privacy computing model, which adaptively selects and combines corresponding privacy computing methods with business data elements and financial tool elements at the atomic level to achieve element privacy protection, provide end-to-end privacy computing service capabilities, realize element circulation and sharing, and enhance the application effect of financial structure data and financial computing capabilities.

[0065] Optionally, step 104 may specifically include: responding to the business operation instruction corresponding to the business scenario, parsing the operation data and operation process corresponding to the business operation instruction; generating scheduling and assembly information corresponding to the business operation instruction based on the operation data and operation process; and calling the target data element in the business data element and the target tool element in the tool processing element as the target operation element according to the scheduling and assembly information.

[0066] The scheduling and assembly information may include the identifier of the target data element to be assembled, the identifier of the target tool element, the assembly logic of the target data element, and the assembly logic of the target tool element.

[0067] In some embodiments, a scheduling and assembly module can be used to construct a highly efficient and intelligent data and tool scheduling device, employing a hybrid plaintext and ciphertext computing architecture to improve the efficiency and accuracy of business processing in financial scenarios. Based on the atomic element scheduling link process input of the business scenario of the end-to-end privacy computing device, a highly efficient and intelligent data and tool scheduling device is constructed using a graph engine computing model. It supports homomorphic encryption and decryption of tool elements and data elements participating in the computation, and constructs a hybrid plaintext + ciphertext privacy computing architecture based on scheduling instructions. It enables the execution of some computation and processing tasks in plaintext on the side owning the element, according to data ownership relationships, with the computation results converted to ciphertext for use in other computation stages. This ensures that elements can only be decrypted into plaintext on the ownership side, while the rest of the process remains in ciphertext. Furthermore, this tool scheduling device supports process auditing of the entire computation process, with relevant operation records stored using blockchain technology, improving audit efficiency through optimized information storage models. The privacy protection capabilities of this module's computation process are achieved through the element-based governance and privacy-preserving computation modules. This part mainly generates scheduling assembly information (SchedulerAddInfo) based on multiple element-based processing paths, forming the final combined capabilities. All computation processes and atomic capabilities still utilize the first two key devices.

[0068] By constructing a highly efficient and intelligent data and tool scheduling device, relying on element-based governance and computational devices, a hybrid computing architecture of plaintext + ciphertext is formed to ensure computational efficiency and accuracy in financial scenarios, balancing efficiency, privacy, and precision across multiple domains. Capabilities such as scheduling link parameterization, graph computation, plaintext-ciphertext hybridization, ownership identification, and process auditing ensure the operational efficiency of the entire privacy computing device. The highly efficient and intelligent data and tool scheduling device, based on the execution plan, acquires corresponding business data elements and financial tool elements, generates scheduling assembly information, and creates the final combined capabilities, realizing a divide-and-conquer-combination privacy computing overall solution, achieving a closed-loop flow from business -> atomic capabilities -> business. This addresses issues such as the lack of comprehensive data protection solutions in related technologies, the inability to simultaneously meet the privacy protection requirements of data models, and the inability to perform privacy protection for element-based models.

[0069] Compared with related technologies, this embodiment can extract business data items from raw business data and tool processing items from business processing flow. It decomposes and maps the business processing flow into business data elements and financial tool elements, automatically mapping them according to the original dimensions of existing data and tools, and forming a business scenario atomic element scheduling link flow. This provides automated analysis capabilities for computational processes based on financial scenario inputs. According to the element's form category and scheduling flow, it constructs client-side privacy computing, data anonymity query, and multi-party secure computing to achieve comprehensive protection of element information. Specifically, this includes providing client software with lightweight computing storage and lightweight privacy protection capabilities, providing localized privacy computing and data storage services for computing participants; providing data anonymity query technical service capabilities based on an upgraded secret sharing system to ensure data participates in computing in an efficient and secure manner; and constructing multi-party secure computing capabilities based on secret sharing to enhance the privacy protection capabilities of business elements and achieve fully encrypted computation of business elements.

[0070] Based on the above Figure 1 The specific implementation of the method shown in this embodiment provides a privacy protection device, such as... Figure 7 As shown, the device includes: an acquisition module 31, a routing module 32, a determination module 33, an encryption module 34, and a processing module 35; Module 31 is used to acquire raw business data and corresponding business tools in the business scenario. The routing module 32 is used to decompose the business processing flow of the original business data and business tools through preset element-based processing routing rules, so as to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools. The determination module 33 is used to determine the target operation element corresponding to the business operation instruction from the business data elements and tool processing elements in response to the business operation instruction corresponding to the business scenario. The encryption module 34 is used to determine the privacy operation corresponding to the target operation element according to the element category and scheduling process of the target operation element, and to encrypt the target operation element by using the privacy operation to assemble the encrypted target processing flow. Processing module 35 is used to generate business processing results corresponding to business operation instructions based on the encrypted target processing flow.

[0071] In some examples, the routing module 32 is specifically configured to configure preset element-based processing routing rules based on the data characteristics of the original business data and the tool processing characteristics of the business processing flow; based on the preset element-based processing routing rules, match the business keywords corresponding to the original business data and the business processing flow, determine the routing path corresponding to the original business data and the business processing flow based on the keyword matching results; and decompose the original business data and the business processing flow according to the routing path corresponding to the original business data and the business processing flow to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools.

[0072] In some examples, the routing module 32 is specifically configured to: identify business keywords corresponding to the original business data and the process data of the business processing flow; generate rule-based routing steps corresponding to the original business data and the business processing flow based on the matching degree between preset element-based processing routing rules and business keywords, as well as the complexity of the business keywords; convert the original business data and process data into embedded vectors using gated routing, determine at least one processing path corresponding to the original business data and process data based on the compatibility matching degree between the embedded vectors and preset gated processing paths in the gated routing, and perform weighted aggregation on the at least one processing path to determine the gated routing steps corresponding to the original business data and process data; identify the expected execution target corresponding to the original business data and process data using an intent recognition model, and generate intent recognition processing steps based on the expected execution target; perform rule reinforcement learning based on the rule-based routing processing steps, gated routing processing steps, and intent recognition processing steps using a reinforcement learning model to generate reinforcement learning processing steps; and determine the routing path corresponding to the original business data and process data based on at least one of the rule-based routing processing steps, gated routing steps, intent recognition processing steps, and reinforcement learning processing steps.

[0073] In some examples, the routing module 32 is specifically configured to use a data extraction tool configured with a preset entry point to extract business data items in batches from the original business data during the transmission of the original business data based on the feature definition tool class and the scheduling task tool class, and to perform feature definition and task scheduling processing on the business data items to generate business data elements after the business data items have been transformed into features.

[0074] In some examples, the routing module 32 is specifically configured to use a data extraction tool to classify the business process according to the tool type during the execution of the business process of the business tool, and obtain the classified business process data; based on the element definition tool class and the scheduling task tool class, it extracts the tool processing items in batches from the business process data, and performs element definition and task scheduling processing on the tool processing items to generate tool processing elements after the tool processing item element transformation.

[0075] In some examples, module 33 is specifically configured to respond to business operation instructions corresponding to business scenarios, parse the operation data and operation process corresponding to the business operation instructions; generate scheduling and assembly information corresponding to the business operation instructions based on the operation data and operation process; and, according to the scheduling and assembly information, call the target data element in the business data elements and the target tool element in the tool processing elements as the target operation element.

[0076] In some examples, the acquisition module 31 is specifically configured to acquire raw business data from different sources in the business scenario through the raw data layer; identify the encryption status corresponding to the raw business data; determine the target transmission channel corresponding to the raw business data based on the encryption status corresponding to the raw business data; transmit the raw business data to the basic data layer through the message middleware using the target component corresponding to the target transmission channel; and clean and classify the raw business data through the basic data layer to obtain the classified and stored raw business data.

[0077] Based on the above, Figure 1 Accordingly, this embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. Figure 1 The method shown.

[0078] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0079] Based on the above, Figure 1 The method shown, and Figure 7 To achieve the above objectives, the present application also provides an electronic device, comprising a storage medium and a processor; the storage medium for storing a computer program; and the processor for executing the computer program to implement the above-described virtual device embodiments. Figure 1 The method shown.

[0080] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0081] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0082] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0083] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms, or it can be implemented by hardware. This application can extract business data items from raw business data, extract tool processing items from business processing flow, decompose and map the business processing flow into business data elements and financial tool elements, perform automated mapping according to the original dimensions of existing data and tools, and form a business scenario atomic element scheduling link flow, providing automated analysis capabilities for the calculation process based on financial scenario input; according to the form category and scheduling flow of elements, it constructs client terminal privacy computing, data anonymity query and multi-party secure computing, to achieve comprehensive protection of element information. Specifically, it includes providing client software with lightweight computing storage and lightweight privacy protection capabilities, providing localized privacy computing and data storage services for computing participants; based on an upgraded version of secret sharing, it realizes the technical service capability of data anonymity query, ensuring that data participates in computing in an efficient and secure mode; it constructs multi-party secure computing capabilities based on secret sharing, enhances the privacy protection capability of business elements, and realizes fully encrypted computing of business elements.

[0084] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0085] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A privacy protection method, characterized in that, include: Obtain the original business data in the business scenario and the business tools corresponding to the original business data. The business tools are the business processing models used by the business organizations in the current business scenario. The business processing models include multiple calculation steps determined based on business data and custom business rules. By decomposing the original business data and the business processing flow of the business tools through preset element-based processing routing rules, business data elements corresponding to the original business data and tool processing elements corresponding to the business tools are obtained, including: Configure the preset element-based processing routing rules based on the data characteristics of the original business data and the tool processing characteristics of the business processing flow; Based on the preset element-based processing routing rules, the original business data and the business keywords corresponding to the business processing flow are matched, and the routing path corresponding to the original business data and the business processing flow is determined according to the keyword matching results. According to the routing path corresponding to the original business data and the business processing flow, the original business data and the business processing flow are decomposed to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools. In response to the business operation instruction corresponding to the business scenario, the target operation element corresponding to the business operation instruction is determined from the business data elements and the tool processing elements; Based on the element category and scheduling process of the target operation element, the privacy operation corresponding to the target operation element is determined, and the target operation element is encrypted using the privacy operation to assemble the encrypted target processing flow. Based on the encrypted target processing flow, the business processing result corresponding to the business operation instruction is generated. The step of matching the original business data and the business keywords corresponding to the business processing flow based on the preset element-based processing routing rules, and determining the routing path corresponding to the original business data and the business processing flow based on the keyword matching results, includes: Identify the business keywords corresponding to the original business data and the process data of the business processing flow, respectively; Based on the matching degree between the preset element-based processing routing rules and the business keywords, and the complexity of the business keywords, rule-based routing processing steps corresponding to the original business data and the business processing flow are generated using rule-based routing. The original business data and the process data are converted into embedded vectors using gated routing. Based on the compatibility matching degree between the embedded vectors and the preset gated processing paths in the gated routing, at least one processing path corresponding to the original business data and the process data is determined according to the compatibility matching degree. The at least one processing path is weighted and aggregated to determine the gated routing processing steps corresponding to the original business data and the process data. Using an intent recognition model, identify the expected execution targets corresponding to the original business data and the process data, and generate intent recognition and processing steps based on the expected execution targets; Using a reinforcement learning model, reinforcement learning is performed on the rule-based routing processing steps, gated routing processing steps, and intent recognition processing steps to generate reinforcement learning processing steps. Based on at least one of the rule-based routing processing steps, gating routing processing steps, intent recognition processing steps, and reinforcement learning processing steps, the routing paths corresponding to the original business data and the process data are determined respectively.

2. The privacy protection method according to claim 1, characterized in that, The step of decomposing the original business data and the business processing flow to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools includes: Using a data extraction tool configured with preset entry points, during the transmission of the original business data, business data items are extracted in batches based on the element definition tool class and the scheduling task tool class. The business data items are then processed for element definition and task scheduling to generate business data elements after the business data items have been transformed into elements.

3. The privacy protection method according to claim 2, characterized in that, The step of decomposing the original business data and the business processing flow to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tools also includes: Using the data extraction tool, during the execution of the business process of the business tool, the business process is classified according to the tool type, and the classified business process data is obtained. Based on the element definition tool class and the scheduling task tool class, the tool processing items in the business process data are extracted in batches, and the tool processing items are processed by element definition and task scheduling to generate the tool processing elements after the tool processing items are transformed into elements.

4. The privacy protection method according to claim 1, characterized in that, The step of responding to a business operation instruction corresponding to the business scenario by determining the target operation element corresponding to the business operation instruction from the business data elements and the tool processing elements includes: In response to the business operation instructions corresponding to the business scenario, the operation data and operation process corresponding to the business operation instructions are parsed. Based on the operation data and operation process, the scheduling and assembly information corresponding to the business operation instruction is generated; Based on the scheduling and assembly information, the target data element in the business data element and the target tool element in the tool processing element are invoked as the target operation element.

5. The privacy protection method according to claim 1, characterized in that, The acquisition of raw business data in the business scenario includes: The raw business data from different sources in the business scenario is obtained through the raw data layer; Identify the encryption state corresponding to the original business data, and determine the target transmission channel corresponding to the original business data based on the encryption state corresponding to the original business data. Using the target component corresponding to the target transmission channel, the original business data is transmitted to the basic data layer via message middleware; The original business data is cleaned and classified by the basic data layer to obtain the classified and stored original business data.

6. A privacy protection device, characterized in that, include: The acquisition module is used to acquire the original business data in the business scenario and the business tools corresponding to the original business data. The business tools are the business processing models used by the business organization in the current business scenario. The business processing models include multiple calculation steps determined based on business data and custom business rules. The routing module is used to decompose the original business data and the business processing flow of the business tool according to preset element-based processing routing rules, to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tool. This includes: configuring the preset element-based processing routing rules based on the data characteristics of the original business data and the tool processing characteristics of the business processing flow; matching business keywords corresponding to the original business data and the business processing flow based on the preset element-based processing routing rules, and determining the routing path corresponding to the original business data and the business processing flow based on the keyword matching results; decomposing the original business data and the business processing flow according to the routing path corresponding to the original business data and the business processing flow, to obtain the business data elements corresponding to the original business data and the tool processing elements corresponding to the business tool; the step of matching business keywords corresponding to the original business data and the business processing flow based on the preset element-based processing routing rules, and determining the routing path corresponding to the original business data and the business processing flow based on the keyword matching results, includes: identifying the business keywords corresponding to the process data of the original business data and the business processing flow respectively. The process involves: using rule-based routing to generate rule-based routing steps corresponding to the original business data and the business processing flow, based on the matching degree between the preset element-based processing routing rules and the business keywords, and the complexity of the business keywords; using gated routing to convert the original business data and the process data into embedding vectors, determining at least one processing path corresponding to the original business data and the process data based on the compatibility matching degree between the embedding vectors and the preset gated processing paths in the gated routing, weighted aggregation of the at least one processing path, and determining the gated routing steps corresponding to the original business data and the process data; using an intent recognition model to identify the expected execution target corresponding to the original business data and the process data, and generating intent recognition processing steps based on the expected execution target; using a reinforcement learning model to perform rule reinforcement learning based on the rule-based routing processing steps, gated routing processing steps, and intent recognition processing steps, and generating reinforcement learning processing steps; and determining the routing paths corresponding to the original business data and the process data based on at least one of the rule-based routing processing steps, gated routing processing steps, intent recognition processing steps, and reinforcement learning processing steps. The determination module is used to determine the target operation element corresponding to the business operation instruction from the business data elements and the tool processing elements in response to the business operation instruction corresponding to the business scenario. An encryption module is used to determine the privacy operation corresponding to the target operation element based on the element category and scheduling process of the target operation element, and to encrypt the target operation element using the privacy operation to assemble the encrypted target processing flow. The processing module is used to generate the business processing result corresponding to the business operation instruction based on the encrypted target processing flow.

7. An electronic device, characterized in that, include: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

Images