Enterprise debtor cross-domain portrait feature extraction method and system in combination with multi-party security computing

Abstract

The invention relates to the technical field of data processing, in particular to an enterprise debtor cross-domain portrait feature extraction method and system combined with multi-party security computing, and the method comprises the steps: obtaining enterprise debtor original data locally stored by participants, connecting the participants to a multi-party security computing system through a security communication channel, and carrying out the cross-domain portrait feature extraction of the multi-party security computing system; the method comprises the following steps of: forming an encryption key pair by a student based on a threshold password, distributing a public key to each participant, keeping a private key by an authorization result user, encrypting original data of each participant by using the public key to obtain encrypted enterprise debtor data, and scoring the importance degree of the encrypted data to obtain an encrypted enterprise debtor; the data with the score higher than a preset threshold value is preferentially selected to execute multi-party security calculation, a multi-dimensional ciphertext portrait feature vector is formed through integration based on a ciphertext calculation result output by a multi-party security calculation protocol, the multi-dimensional ciphertext portrait feature vector is output to an authorization result user, and plaintext portrait features are obtained through decryption by a private key. And efficient extraction and integration of cross-domain features are realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation. Background Technology

[0002] In fields such as financial risk control and corporate credit assessment, cross-institutional data collaborative analysis is a core means of accurately profiling corporate debtors. The fusion of heterogeneous data from multiple sources, including tax, banking, power grid, and business registration, can effectively reveal a company's operating status and debt repayment ability. However, traditional data collection models are difficult to implement in compliance with regulations due to data privacy protection laws and data silos between institutions. Privacy-preserving computation technologies, such as multi-party secure computation (MPC), achieve cross-domain feature joint extraction through a mechanism where data is usable but not visible, becoming a key path to resolving the conflict between data sharing and privacy protection.

[0003] However, as the dimensions of enterprise profiling continue to expand and the scale of participating institutions grows, enterprise profiling involves a large number of nonlinear operations, such as division in tax-to-sales ratio calculation and logistic regression in risk scoring models. Homomorphic encryption requires polynomial approximation or interactive protocol conversion for operations such as division and square root, and the computational complexity increases exponentially with the data scale. For example, a pilot project of a bank showed that when processing the tax-to-sales ratio features of 100,000 enterprises, the calculation latency of a single feature soared from seconds to tens of minutes as the number of participants increased, far exceeding the real-time response threshold required by financial risk control scenarios.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a method and system for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation. This aims to solve the technical problem that in cross-institutional data collaborative analysis in fields such as financial risk control, the computational complexity of privacy-preserving computation technologies such as homomorphic encryption increases exponentially with the data scale when handling nonlinear complex operations such as division and logistic regression, resulting in excessive latency and insufficient effectiveness in large-scale data processing.

[0006] To achieve the above objectives, this invention provides a method for cross-domain profiling feature extraction of corporate debtors that combines multi-party secure computation, the method comprising: Obtain the original corporate debtor data stored in the local database of each participating party, wherein each participating party includes commercial banks, tax bureaus and power grid companies, and the original corporate debtor data includes basic corporate information, financial account data, tax data and energy consumption data; Each participant is connected to the multi-party secure computing system through a secure communication channel. When the central scheduler sends a feature generation computing task instruction to each participant, the system generates the encryption key pair required for the multi-party secure computing protocol based on threshold cryptography or a trusted third party. The public key is distributed to each participant, and the private key is kept by the authorized result user. The original data of the corporate debtors of each participating party are encrypted using a public key to obtain encrypted corporate debtor data. The importance of the encrypted corporate debtor data is scored, and the encrypted corporate debtor data with an importance score greater than a preset score threshold is preferentially selected to perform secure computation using the corresponding multi-party secure computation technology combination. Based on the encrypted computation results output by the multi-party secure computation protocol, an encrypted profile feature vector of the corporate debtor containing multi-dimensional features is formed. The encrypted profile feature vector of the corporate debtor is then output to the authorized result user, so that the authorized result user can obtain the plaintext profile features of the corporate debtor with multi-dimensional features by decrypting with a private key.

[0007] Optionally, the step of preferentially selecting a corresponding multi-party secure computation technology combination to perform secure computation on encrypted corporate debtor data whose importance score is greater than a preset scoring threshold includes: The system reads encrypted corporate debtor data whose importance score is greater than a preset score threshold in blocking mode, and performs secure computation on the encrypted corporate debtor data read in blocking mode by preferentially selecting the corresponding combination of multi-party secure computation techniques.

[0008] Optionally, after prioritizing the selection of corresponding multi-party secure computation technology combinations to perform secure computation on encrypted corporate debtor data whose importance scores are greater than a preset scoring threshold, the method further includes: Read encrypted corporate debtor data whose importance score is less than a preset score threshold using non-blocking mode; After the encrypted corporate debtor data read in the blocking mode is subjected to security computation using a combination of corresponding multi-party secure computation techniques, the encrypted corporate debtor data read in the non-blocking mode is subjected to security computation using a combination of corresponding multi-party secure computation techniques.

[0009] Optionally, the step of preferentially selecting a corresponding multi-party secure computation technology combination to perform secure computation on encrypted corporate debtor data whose importance score is greater than a preset scoring threshold includes: The computational complexity, real-time requirements, and current network conditions of encrypted corporate debtor data whose importance scores exceed a preset scoring threshold are determined. The allocation ratio of computing resources between edge devices and cloud servers is determined based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data. Based on the allocation ratio of computing resources between the edge devices and cloud servers, priority is given to selecting the corresponding combination of multi-party secure computing technologies to perform secure computing.

[0010] Optionally, determining the allocation ratio of edge devices and cloud server computing resources based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data includes: Based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data, the encrypted corporate debtor data is divided into multiple sub-task data packets; The allocation ratio of computing resources between edge devices and cloud servers is determined based on the subtask data packets.

[0011] Optionally, the encrypted corporate debtor data is split into multiple sub-task data packets based on the computational complexity, real-time requirements, and current network conditions, including: Based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data, the encrypted corporate debtor data is divided into multiple functional modules according to function or logic. Determine the dependencies and data flow relationships between the various functional modules; Multiple subtask data packets are determined based on the dependencies and data flow relationships between the functional modules.

[0012] Optionally, determining the allocation ratio of edge device and cloud server computing resources based on the subtask data packet includes: Based on the characteristics and resource requirements of the subtask data packets, determine the corresponding subtasks that the encrypted corporate debtor data node and its extended modules can undertake; The data calculation execution order is determined based on the corresponding sub-tasks that the encrypted corporate debtor data node and its extended modules can undertake. The allocation ratio of computing resources between edge devices and cloud servers is determined based on the execution order of the data.

[0013] Optionally, the feature vector of the corporate debtor encrypted profile with multi-dimensional features includes cross-features of taxation and finance, features of energy consumption and operating status, features of cross-institutional financial attributes, and features of joint risk assessment.

[0014] Optionally, after enabling the authorized user to decrypt the multi-dimensional features of the corporate debtor's plaintext profile using a private key, the process further includes: The risk interpretation is performed on the plaintext profile features of the corporate debtors with the aforementioned multi-dimensional characteristics, and risk control decision suggestions are output. The risk interpretation is generated based on the correlation analysis of the multi-dimensional features, including risk level, explanation basis and business suggestions.

[0015] Furthermore, to achieve the above objectives, the present invention also provides a cross-domain profile feature extraction system for corporate debtors that incorporates multi-party secure computation. The system includes a cross-domain profile feature extraction device for corporate debtors that incorporates multi-party secure computation. The device includes a memory, a processor, and a cross-domain profile feature extraction program for corporate debtors that incorporates multi-party secure computation, stored in the memory and executable on the processor. The cross-domain profile feature extraction program for corporate debtors that incorporates multi-party secure computation is configured to implement the steps of the cross-domain profile feature extraction method for corporate debtors that incorporates multi-party secure computation as described above.

[0016] This invention provides a method and system for cross-domain profiling feature extraction of corporate debtors by combining multi-party secure computation. The method includes: acquiring raw corporate debtor data stored in local databases by each participating party, wherein the participating parties include commercial banks, tax bureaus, and power grid companies, and the raw corporate debtor data includes basic corporate information, financial account data, tax data, and energy consumption data; connecting each participating party to a multi-party secure computation system through a secure communication channel; and, upon receiving feature generation and computation task instructions from a central scheduler, generating encryption key pairs required for the multi-party secure computation protocol based on threshold cryptography or a trusted third party, distributing the public key to each participating party, and private key to the private key. The key is kept by the authorized result user. The original data of the corporate debtors of each participating party is encrypted using the public key to obtain encrypted corporate debtor data. The importance of this encrypted corporate debtor data is scored, and encrypted corporate debtor data with an importance score greater than a preset threshold are preferentially selected for secure computation using corresponding multi-party computation techniques. Based on the ciphertext computation results output by the multi-party secure computation protocol, a ciphertext profile feature vector of corporate debtors containing multi-dimensional features is integrated and output to the authorized result user, enabling the authorized result user to decrypt using their private key to obtain the plaintext profile features of the corporate debtors with multi-dimensional features. This invention filters encrypted data through an importance scoring mechanism, prioritizing secure computation only on key data with scores greater than a preset threshold, avoiding indiscriminate complex calculations on all data simultaneously. Key data, such as the tax-to-sales ratio in tax data and the flow fluctuation characteristics in financial accounts, ensures the priority of secure computation for core sensitive features through data grading. This improves efficiency without compromising privacy protection, achieving the dual goals of data usability without visibility and efficient and scalable computation. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of the cross-domain profile feature extraction device for corporate debtors that combines multi-party secure computation with the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating an embodiment of the cross-domain profiling feature extraction method for corporate debtors that incorporates multi-party secure computation according to the present invention. Figure 3 This is a schematic diagram of the framework of an embodiment of the corporate debtor cross-domain profiling feature extraction system that combines multi-party secure computation according to the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a cross-domain profiling feature extraction device for corporate debtors that combines multi-party secure computation with the hardware operating environment involved in the embodiments of the present invention.

[0021] like Figure 1 As shown, the enterprise debtor cross-domain profile feature extraction device combining multi-party secure computation may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen, and optionally, it may also include a standard wired interface or a wireless interface. In this invention, the wired interface of the user interface 1003 may be a USB interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0022] Those skilled in the art will understand that Figure 1The structure shown does not constitute a limitation on the cross-domain profiling feature extraction device for corporate debtors that incorporates multi-party secure computation. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0023] like Figure 1 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a cross-domain profile feature extraction program for corporate debtors that incorporates multi-party secure computation.

[0024] exist Figure 1 In the enterprise debtor cross-domain profile feature extraction device combining multi-party secure computation shown, the network interface 1004 is mainly used to connect to the backend server and communicate with the backend server; the user interface 1003 is mainly used to connect to peripheral devices; the enterprise debtor cross-domain profile feature extraction device combining multi-party secure computation calls the enterprise debtor cross-domain profile feature extraction program combining multi-party secure computation stored in the memory 1005 through the processor 1001, and executes the enterprise debtor cross-domain profile feature extraction method combining multi-party secure computation provided in this embodiment of the invention.

[0025] Based on the above hardware structure, an embodiment of the present invention is proposed, which combines multi-party secure computation to extract cross-domain profile features of corporate debtors.

[0026] Reference Figure 2 , Figure 2 This is a flowchart illustrating an embodiment of the cross-domain profiling feature extraction method for corporate debtors combining multi-party secure computation according to the present invention. In this embodiment, the cross-domain profiling feature extraction method for corporate debtors combining multi-party secure computation includes the following steps: S10: Obtain the original corporate debtor data stored in the local database of each participating party, wherein each participating party includes commercial banks, tax bureaus and power grid companies, and the original corporate debtor data includes basic corporate information, financial account data, tax data and energy consumption data.

[0027] It should be noted that "participants" refer to various independent entities or institutions that participate in data sharing, computation, or collaboration within a specific scenario, project, or system. In the context of cross-institutional data collaborative analysis, participants can be institutions or organizations that possess relevant data and are willing to share data and conduct joint computations while protecting privacy. Specifically, commercial banks provide corporate debtors' financial account data, such as corporate account statements, loan records, and repayment status, to assess the company's liquidity and debt repayment ability. Tax bureaus provide corporate debtors' tax data, such as tax payments, value-added tax, and income tax, to analyze the company's operating conditions and profitability. Power grid companies provide corporate debtors' energy consumption data, such as electricity consumption and electricity bill payment records, to reflect the company's production scale and operational stability.

[0028] Raw debtor data refers to the unprocessed, most basic set of data provided by various participating parties when assessing a debtor's creditworthiness, solvency, and operational risks. This data originates directly from the company's daily operations and management systems, encompassing multiple dimensions such as basic information, financial data, tax data, and energy consumption data. It forms the underlying data foundation for extracting debtor profile features and conducting risk analysis. Basic company information includes the company name, registered address, legal representative, establishment date, and business scope, used to identify the company's identity and background. Financial account data includes bank account transaction records, loan information, and repayment records, used to analyze the company's liquidity and creditworthiness. Tax data includes the company's tax records, tax types, and tax amounts, used to assess the company's operational compliance and profitability. Energy consumption data includes the company's electricity consumption and electricity bill payment records, used to infer the company's production scale and operational status.

[0029] In practice, commercial banks extract financial account data from corporate debtors through their internal systems, including corporate account statements, loan records, and repayment details. They ensure the data format conforms to the input requirements of a multi-party secure computation system. This internal system can be a core banking system or a credit management system, and the data format can be CSV or JSON. The tax bureau extracts tax data from corporate debtors through its tax system, including tax records, tax types, and tax amounts, and anonymizes sensitive fields. This tax system can be the Golden Tax System Phase III, and the sensitive field can be the corporate taxpayer identification number. The power grid company extracts energy consumption data from corporate debtors through its electricity metering system, including electricity consumption and payment records, and standardizes the data, such as unifying the unit of measurement to kilowatt-hours. Each participating party stores the extracted raw data in a local encrypted database to ensure data security during transmission and storage. The data undergoes cleaning and preprocessing, including removing duplicate values, filling in missing values, and standardizing data formats to improve data quality and computational efficiency. Data formats include date formats and currency units.

[0030] S20: Connect each participant to the multi-party secure computing system through a secure communication channel. When the central scheduler sends a feature generation computing task instruction to each participant, it generates the encryption key pair required for the multi-party secure computing protocol based on threshold cryptography or a trusted third party. The public key is distributed to each participant, and the private key is kept by the authorized result user.

[0031] It should be noted that a secure communication channel refers to a dedicated data transmission channel constructed using encryption technology. This ensures the confidentiality, integrity, and non-repudiation of data transmitted between the participating parties and the multi-party secure computing system, preventing data theft, tampering, or forgery. Encryption technologies include SSL / TLS protocols and VPNs. A multi-party secure computing system is a technology platform that supports multiple data holders in collaboratively completing joint computations without directly sharing the original data. Its core function is to achieve data usability without visibility through cryptographic protocols, ensuring the privacy of all parties' data during the computation process. These cryptographic protocols include homomorphic encryption, secret sharing, and obfuscated circuits. The central scheduler is responsible for coordinating the entire process of multi-party secure computing, including task allocation, resource scheduling, key management, and computing node monitoring. Its role is to uniformly manage the computing tasks of all participating parties, ensuring the orderly and efficient collaboration among them.

[0032] Threshold cryptography is a key sharding technique that divides the private key into multiple subkeys (shards) and distributes them to different participants. The complete private key can only be reconstructed when at least a preset threshold number of subkeys are collected. For example, using the (t,n) threshold scheme, at least t of the n participants must cooperate to recover the private key, improving key security and preventing a single node from possessing the complete private key. A trusted third party, in traditional multi-party secure computation, is an entity that acts as a neutral body responsible for generating and managing keys and coordinating the computation process. It relies on the assumption of absolute trust from all parties, but in practical applications, single points of failure or trust risks can be introduced. Therefore, threshold cryptography and this scheme are two alternative key generation methods. An encryption key pair is an asymmetric encryption key combination consisting of a public key and a private key: the public key can be publicly distributed and used to encrypt data or verify signatures; anyone can use the public key to encrypt data, but it is impossible to deduce the private key from the public key. The private key is strictly confidential and held only by authorized parties, used to decrypt data encrypted with the public key or generate signatures. Authorized result users refer to entities authorized by all participating parties that have the right to receive and decrypt the final calculation results, such as commercial banks and other financial institutions. Their responsibility is to use the private key to decrypt the plaintext profile features after the multi-party secure calculation is completed, and then use them for business scenarios such as credit approval and risk assessment.

[0033] In practice, each participating party connects to the multi-party secure computing system via dedicated lines or encrypted networks. SSL / TLS protocols are used for end-to-end encryption of transmitted data, ensuring that data is encrypted into ciphertext during transmission and cannot be cracked even if intercepted. Upon access, participating parties must authenticate themselves using digital certificates or dynamic tokens to prevent unauthorized organizations from impersonating them. For example, tax authorities must provide government-specific certificates, and commercial banks must be certified by financial certification authorities. Each participating party registers its data type and computing capabilities with the central scheduler so that the central scheduler can subsequently allocate suitable computing tasks. Data types include tax data and financial data, and computing capabilities include whether homomorphic encryption operations are supported.

[0034] When a user of the authorized results needs to generate a profile of a corporate debtor, they submit a feature generation request to the central scheduler, such as "calculate the debt repayment capacity feature vector of company A." The central scheduler parses the required data dimensions and calculation logic based on the request, generating specific feature generation calculation task instructions. Data dimensions include financial account data, tax data, and energy data, and the calculation logic is trained using a logistic regression model. The central scheduler broadcasts the task instructions to all relevant participants through a secure communication channel. The instructions include a unique task identifier, required data fields, calculation protocol type, and data encryption requirements. These relevant participants include the tax bureau, power grid company, and corresponding commercial bank involved in the company's data. The calculation protocol type may be threshold cryptography or a trusted third-party model.

[0035] When using threshold cryptography to generate key pairs, the central scheduler initiates a threshold cryptography algorithm, such as Shamir's Secret Sharing, to randomly split the private key into n fragments, where n is the number of participants. Each participant is assigned a unique subkey (fragment), and a minimum threshold t is set to recover the private key (e.g., t=2, meaning at least two participants must collaborate to reconstruct the private key). Each participant sends their subkey to the central scheduler via a secure channel. The central scheduler then synthesizes a global public key based on the threshold algorithm. This public key can be used to encrypt data but cannot be used to deduce the private key fragments. When using a trusted third party to generate key pairs, the trusted third party acts as the key generation center, generating asymmetric encryption key pairs. The public key is distributed to all participants, and the private key is sent directly by the trusted third party to authorized users of the results. The trusted third party can be an industry regulatory body or a certified neutral institution, and digital signatures ensure the integrity of key transmission. It's important to note that the trusted third-party model relies on the trust of all parties and is suitable for scenarios with strict regulatory requirements and strong third-party credibility, such as government data sharing platforms.

[0036] It should be understood that, regardless of the mode used, the generated public key is securely distributed to all participants through a central scheduler. Each participant uses the public key to encrypt their local raw data, such as encrypting corporate tax data into ciphertext. If threshold cryptography is used, the private key consists of fragments of subkeys held by each participant, eliminating the need for centralized storage. Only during decryption does the authorized result user collect at least t fragments and reconstruct the complete private key, ensuring secure fragment transmission. If a trusted third party is used, the private key is directly managed by the authorized result user and stored in a hardware security module or encrypted database, ensuring multi-factor authentication for private key access. Upon receiving the public key, each participant can verify its integrity through hash verification or digital signatures to prevent man-in-the-middle attacks that could replace malicious public keys. After receiving "key reception confirmation" from all participants, the central scheduler marks the task as "ready" and proceeds to the next stage of data encryption and secure computation.

[0037] S30: Use the public key to encrypt the original data of the corporate debtors of each participating party to obtain encrypted corporate debtor data, and score the importance of the encrypted corporate debtor data. For encrypted corporate debtor data with an importance score greater than a preset score threshold, select the corresponding multi-party secure computation technology combination to perform secure computation.

[0038] It's important to note that importance scoring refers to quantifying and assigning a score to each field or subset of raw data from a corporate debtor based on its impact on the final computational objective. A higher score indicates a greater contribution of the data to feature generation or model training; for example, the "tax amount" field in tax data might be more important than the "corporate registered address." Scoring criteria typically include the data's business relevance, information gain ratio, and historical model contribution. A preset scoring threshold is a critical value for importance pre-defined based on business needs or historical experience. When the importance score exceeds this threshold, the data is considered "high-value data," requiring priority to employ more secure or computationally accurate multi-party computation techniques. "Low-value data" below the threshold can utilize lightweight techniques to improve efficiency. Threshold settings must align with industry standards; for instance, in financial risk control, the threshold for tax data is typically higher than for basic information. Multi-party computation techniques are combined based on data sensitivity, computational complexity, and importance, using various secure computation techniques such as homomorphic encryption, secret sharing, obfuscated circuits, and federated learning. For example, highly important data, such as tax data and loan balances, uses a combination of "obfuscated circuits + homomorphic encryption" to ensure privacy and computational accuracy; less important data, such as a company's business scope and establishment date, uses a combination of "secret sharing + lightweight encryption" to reduce computational overhead.

[0039] In practice, each participating party extracts raw debtor data from its local database and splits it by field or data block, such as treating "loan balance," "tax payment," and "average monthly electricity consumption" as independent data blocks. Participants use the public key distributed during the process to encrypt each data block, generating encrypted debtor data. For example, a commercial bank encrypts the loan balance field of company A as ciphertext C1, and the tax bureau encrypts the tax payment as ciphertext C2. The encrypted data retains metadata tags, such as data type, participating party, and associated company ID, facilitating subsequent scoring and matching. A central scheduler or independent scoring module generates a score of 0-10 for each encrypted data block based on preset rules or machine learning models, such as a 9-point score for "tax payment" and a 3-point score for "business scope." The scoring process must record the basis for weight allocation, such as 60% for business expert rules and 40% for historical data training, ensuring interpretability and compliance.

[0040] Compare the score of each encrypted data block with a preset scoring threshold (e.g., a threshold of 6 points): High-value data (score > 6 points): such as "loan balance" and "number of overdue payments" in financial account data, "value-added tax payment" in tax data, and "average monthly electricity consumption" in energy data. Low-value data (score ≤ 6 points): such as "legal representative's name" and "registered phone number" in basic enterprise information, and "account opening time" in financial data. Store high-value data and low-value data in different task queues, with high-value data entering the "priority calculation queue" and low-value data entering the "regular calculation queue".

[0041] The central scheduler groups data and allocates a combination of "obfuscated circuits + homomorphic encryption" to high-value data: first, homomorphic encryption is used to perform addition / multiplication operations on the ciphertext of tax amounts, such as calculating a company's total tax payment for the past 12 months; then, obfuscated circuits are used to verify logical conditions, such as "whether the tax payment has decreased for three consecutive months." Low-value data employs secret sharing technology: the hash value of the company's registered address is fragmented and distributed to multiple participants, and reconstruction is only performed using a threshold when deduplication is needed, avoiding the transmission of the complete plaintext. More computing nodes or higher-priority computing power, such as GPU clusters, are allocated to high-value data tasks to ensure that their secure computation processes are executed first. For example, complex computational tasks involving tax data occupy 80% of system resources, while basic information processing uses only 20%. The central scheduler tracks the computation progress of high-value data in real time. If delays occur, such as excessively long obfuscated circuit computation times, resource expansion or technology degradation is automatically triggered, such as switching from full homomorphic encryption to partial homomorphic encryption, balancing efficiency and security.

[0042] Furthermore, in another embodiment, the step of preferentially selecting corresponding multi-party secure computation technology combinations to perform secure computation on encrypted corporate debtor data whose importance scores are greater than a preset scoring threshold includes: The system reads encrypted corporate debtor data whose importance score is greater than a preset score threshold in blocking mode, and performs secure computation on the encrypted corporate debtor data read in blocking mode by preferentially selecting the corresponding combination of multi-party secure computation techniques.

[0043] It's important to note that blocking mode is a synchronous data reading mechanism: when a program attempts to read specific data (such as highly important encrypted corporate debtor data), it will wait until the data is ready or a timeout occurs, during which time no other tasks will be processed. Unlike non-blocking read mode (which allows read operations to return immediately regardless of data readiness), blocking mode ensures that high-priority data is processed first through exclusive waiting, avoiding delays caused by low-priority tasks preempting resources. It is suitable for scenarios with extremely high real-time and priority requirements and ensures the security of data transmission.

[0044] In practical implementation, when the secure computing module needs to acquire high-value data, it proactively initiates a blocking read request to the high-priority queue. The computing thread enters a blocked state until a data block meeting the criteria is available in the queue—that is, the data has been encrypted and its importance has been assessed. During this period, it does not respond to read requests from the low-priority queue. A blocking timeout is set (e.g., 300ms). If the data is not acquired within the timeout period, thread resources are released to avoid deadlock. This timeout needs to be adjusted based on business tolerance; in financial scenarios, the timeout is typically shorter. After the data is read, its integrity and permissions are immediately verified to prevent the intrusion of illegal data. By reading high-importance data in blocking mode, a strong binding between data priority and computing resources is achieved, ensuring that data critical to business decisions is processed preferentially and securely.

[0045] Furthermore, after prioritizing the use of corresponding multi-party secure computation technology combinations to perform secure computation on encrypted corporate debtor data whose importance scores exceed a preset scoring threshold, the method further includes: Read encrypted corporate debtor data whose importance score is less than a preset score threshold using non-blocking mode; After the encrypted corporate debtor data read in the blocking mode is subjected to security computation using a combination of corresponding multi-party secure computation techniques, the encrypted corporate debtor data read in the non-blocking mode is subjected to security computation using a combination of corresponding multi-party secure computation techniques.

[0046] It's important to note that non-blocking mode reads are an asynchronous data reading mechanism: when a program attempts to read low-priority encrypted corporate debtor data, it can return immediately without waiting for the data to be ready, allowing the thread to continue processing other tasks, such as calculating high-priority data or system monitoring. Once the data is ready, subsequent processing is triggered via a callback mechanism or event notification. Unlike the exclusive waiting of blocking mode, non-blocking mode improves system resource utilization through asynchronous operations. It is suitable for low-priority tasks with low security requirements, such as format verification of basic corporate information, and because it can return immediately without waiting for the data to be ready, it improves data processing speed.

[0047] In another embodiment, the step of preferentially selecting a corresponding multi-party secure computation technology combination to perform secure computation on encrypted corporate debtor data whose importance score is greater than a preset scoring threshold includes: The computational complexity, real-time requirements, and current network conditions of encrypted corporate debtor data whose importance scores exceed a preset scoring threshold are determined. The allocation ratio of computing resources between edge devices and cloud servers is determined based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data. Based on the allocation ratio of computing resources between the edge devices and cloud servers, priority is given to selecting the corresponding combination of multi-party secure computing technologies to perform secure computing.

[0048] It's important to note that computational complexity refers to the algorithmic complexity, data processing volume, and computing power consumption required for secure computation of encrypted corporate debtor data. For example, high complexity involves ciphertext multiplication, division, and nested logical judgments, such as training a risk scoring model under homomorphic encryption; low complexity only requires ciphertext addition, comparison, or simple aggregation, such as counting the number of companies under secret sharing. Real-time requirements are categorized based on the time sensitivity of the computation results in the business scenario, typically into strong real-time and weak real-time requirements. Strong real-time requirements include real-time risk assessment in loan approval; weak real-time requirements include monthly corporate credit report generation. Edge devices, distributed computing nodes deployed near data collection or usage terminals, such as enterprise local servers, financial institution front-end machines, and IoT gateways, possess lightweight computing power and low-latency data processing capabilities, making them suitable for handling low-complexity, high-real-time tasks, avoiding the latency and bandwidth consumption associated with long-distance data transmission to the cloud. The cloud server computing resource allocation ratio dynamically divides the computing power allocation strategy between edge devices and cloud servers according to task requirements. For example, 60% of the edge processing is simple computing and 40% of the cloud processing is complex computing, realizing a collaborative architecture of "lightweight edge processing + deep cloud computing" to balance latency, computing power and security.

[0049] The determination of the allocation ratio of computing resources between edge devices and cloud servers based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data includes: Based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data, the encrypted corporate debtor data is divided into multiple sub-task data packets; The allocation ratio of computing resources between edge devices and cloud servers is determined based on the subtask data packets.

[0050] It should be noted that the subtask data package is a lightweight task unit that breaks down the original encrypted corporate debtor data and its computational tasks according to functional modules, computational steps, and data sensitivity. It can include task descriptions, data parameters, resource requirements, and dependencies. The task description is the specific computational logic, such as "encrypted addition of corporate tax amount" or "logic judgment of loan delinquency count." Data parameters are encrypted sub-data blocks, such as "encrypted tax amount for Q3 2024" or "encrypted number of related companies." Resource requirements include computing power level, network transmission volume, and real-time constraints. Dependencies are upstream and downstream task relationships, such as "subtask B needs to wait for the encrypted result of subtask A." Essentially, it is a minimal encapsulation of task granularity, supporting distributed collaborative processing between edge devices and cloud servers. By splitting encrypted corporate debtor data into subtask data packages with clear resource requirements, and dynamically calculating the allocation ratio based on edge-cloud capability profiles, the solution achieves precise splitting, on-demand scheduling, and collaborative execution of secure computation tasks. This fine-grained resource allocation mechanism leverages the low latency of edge devices while utilizing the powerful computing capabilities of the cloud, providing a distributed solution that balances efficiency, cost, and compliance for secure computing of high-value data. It is particularly suitable for complex business scenarios involving cross-regional and multi-organizational collaboration.

[0051] Specifically, the process of splitting the encrypted corporate debtor data into multiple sub-task data packets based on the computational complexity, real-time requirements, and current network conditions includes: Based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data, the encrypted corporate debtor data is divided into multiple functional modules according to function or logic. Determine the dependencies and data flow relationships between the various functional modules; Multiple subtask data packets are determined based on the dependencies and data flow relationships between the functional modules.

[0052] It should be noted that a functional module refers to the smallest executable unit that breaks down the complete computation task of encrypting corporate debtor data according to business function attributes or computational logic independence. It possesses the characteristics of single responsibility, clear input and output, and technical adaptability. Single responsibility means implementing only a specific function, such as "encrypted verification of basic corporate information" or "encrypted calculation of loan overdue days." Clear input and output means including standardized encrypted input parameters and encrypted output results, such as "encrypted corporate code" or "encrypted loan contract number," and encrypted output results such as "verification passed / failed" or "encrypted overdue days." Technical adaptability means being able to independently match the computing technologies of edge devices or cloud servers; for example, lightweight encryption modules adapt to the edge, while complex algorithm modules adapt to the cloud. Essentially, it is an atomic encapsulation of business logic, providing a basic unit for subsequent dependency analysis and resource allocation. Dependencies refer to the execution order constraints between functional modules, including temporal dependencies, data dependencies, and resource dependencies. Temporal dependencies mean that a subsequent module can only start after the preceding module outputs its result; for example, the "Risk Scoring Model Calculation" module depends on the output of the "Feature Ciphertext Aggregation" module. Data dependencies mean that the input parameters of a subsequent module must be provided by the output of the preceding module; for example, the "Tax Amount Percentage Ciphertext Calculation" module depends on the outputs of the "Tax Amount Ciphertext" and "Operating Revenue Ciphertext" modules. Resource dependencies refer to the mutual exclusion relationships when multiple modules compete for the same computing resources, such as the queuing order when two modules both need to call the SM4 hardware encryption unit of an edge device. Dependency definitions ensure that subtask data packets conform to the business logic order during distributed execution. Data flow relationships refer to the encrypted transmission path of data between functional modules, including data direction, data format, and data volume. Data direction specifies which module's output the encrypted data is transmitted to, such as "enterprise equity structure encrypted data" flowing from the "business data parsing" module to the "relationship analysis" module. Data format defines the encrypted encoding format of the transmitted data, such as homomorphically encrypted integer ciphertext or secretly shared fragmented ciphertext. Data volume quantifies the size of the encrypted data in a single flow, such as a 10KB encrypted data packet or a 500KB encrypted data packet, serving as the basis for network resource allocation. Data flow relationships directly affect the network transmission efficiency of subtask data packets and the selection of secure computing technologies.

[0053] It should be understood that by breaking down encrypted corporate debtor data into the smallest executable units, i.e., functional modules, according to function / logic, the originally coupled complex computational tasks are transformed into independent and controllable atomic operations. For example, "corporate risk score calculation" is broken down into modules such as "basic data verification," "feature ciphertext extraction," and "model inference," so that each module undertakes only a single computational function. This breaks the traditional black-box task processing mode, realizes the explicit decomposition of computational logic, and facilitates subsequent independent resource adaptation, performance optimization, and security auditing for each module. Unlike the problem of overly coarse scheduling granularity caused by traditional whole-task packaging processing, such as directly scheduling 10GB of ciphertext data to the cloud and ignoring the lightweight processing capabilities of edge devices, this embodiment uses granular and controllable module decomposition, enabling the system to dynamically adjust the size of processing units according to real-time network conditions. When the network is congested, it is split into 10KB small packets for edge processing, and when the network is good, it is merged into 100KB large packets for cloud processing, improving scheduling flexibility by more than 70%.

[0054] Specifically, determining the allocation ratio of edge device and cloud server computing resources based on the subtask data packet includes: Based on the characteristics and resource requirements of the subtask data packets, determine the corresponding subtasks that the encrypted corporate debtor data node and its extended modules can undertake; The data calculation execution order is determined based on the corresponding sub-tasks that the encrypted corporate debtor data node and its extended modules can undertake. The allocation ratio of computing resources between edge devices and cloud servers is determined based on the execution order of the data.

[0055] It should be noted that encrypted corporate debtor data nodes refer to physical or logical computing units with encrypted data processing capabilities. They are the basic execution entities carrying subtask data packets, and can be edge devices deployed locally within an enterprise, such as industrial control computers or smart gateways, or logical computing nodes in a cloud server cluster. They support secure computation of encrypted data, such as homomorphic encryption and secret sharing decryption; they have built-in secure computing engines, such as supporting SM2 / SM3 / SM4 national cryptographic algorithms, and possess security capabilities such as encrypted data transmission and access control, ensuring the security of encrypted corporate debtor data processed within the node; each node has clearly defined upper limits for computing power, storage, and network bandwidth as constraints for resource allocation. Extension modules refer to functional enhancement components that can be dynamically loaded onto data nodes to expand the node's encrypted data processing capabilities, including computing power extension modules, storage extension modules, and network extension modules. Computing power extension modules include GPU accelerator cards and dedicated encryption chips; storage extension modules include distributed caching components and secure database plugins; and network extension modules include 5G communication modules for edge nodes and RDMA network adapters for cloud nodes. The extension module can be dynamically mounted according to the specific needs of the subtask data packet. For example, when handling high-complexity homomorphic encryption tasks, a GPU acceleration module can be temporarily loaded for cloud nodes to achieve flexible adaptation of node capabilities.

[0056] It should be understood that by analyzing the computing power requirements, real-time requirements, and data volume of subtask data packets, they are allocated to the most suitable nodes and expansion modules. Low computing power requirements and tasks with strong real-time demands are processed directly by edge nodes, avoiding network latency from remote calls to the cloud and reducing response time by 70%. Localized compliance tasks are forcibly assigned to local enterprise edge nodes; localized compliance tasks such as "encrypted verification of tax data" must meet the requirement that data does not leave the enterprise's premises. High computing power requirements and batch processing tasks are scheduled to the cloud, utilizing its elastic computing power and distributed storage capabilities, improving computing efficiency by more than 5 times; high computing power requirements and batch processing tasks such as "encrypted modeling of historical default data" require 100 cores for parallel computing. Cross-regional collaborative tasks are uniformly coordinated by cloud nodes, solving the problem of distributed computing power at edge nodes; cross-regional collaborative tasks such as "encrypted analysis of cross-border enterprise relationships" are examples of such tasks.

[0057] Furthermore, extension modules are loaded to address the specific needs of subtasks, overcoming the inherent limitations of node capabilities. For example, when a subtask involves a large number of encrypted matrix multiplications, a GPU acceleration module is temporarily loaded onto the cloud node, increasing the computation speed by 300%, from 500ms / time for CPU processing to 150ms / time for GPU. Similarly, when edge nodes process large batches of encrypted data storage, a distributed cache extension module is dynamically mounted, increasing the encrypted read speed from 10MB / s to 50MB / s, resolving the bottleneck of insufficient edge memory. The execution order is determined based on the dependencies of subtask data packets, avoiding logical chaos in distributed computing. For instance, if a preceding subtask is not completed, a subsequent subtask will not be scheduled for execution. A global execution sequence is generated through a dependency graph traversal algorithm, ensuring that encrypted data flows according to business logic. Preceding subtasks include "basic data encrypted verification packages," and subsequent subtasks include "risk model inference packages." Subsequent nodes only receive the complete output encrypted data from preceding nodes, avoiding computational errors caused by incomplete data fragmentation in traditional distributed systems. The task failure rate is reduced from 5% in traditional solutions to below 0.1%. By reducing unnecessary waiting between nodes through execution sequence planning, for example, for subtasks with no dependencies, such as "enterprise basic information verification package" and "financial statement encrypted parsing package", they can be scheduled to be processed in parallel on different edge nodes, saving 50% of the time compared to serial execution; after the front-end node completes the subtask, it immediately triggers the pre-allocation of resources for the subsequent nodes, such as applying for GPU resources for cloud nodes in advance, so that the task connection delay between nodes is reduced from 100ms in the traditional solution to less than 10ms.

[0058] S40: Based on the ciphertext computation results output by the multi-party secure computation protocol, integrate them to form a corporate debtor ciphertext profile feature vector containing multi-dimensional features, and output the corporate debtor ciphertext profile feature vector to the authorized result user so that the authorized result user can decrypt the corporate debtor plaintext profile features with multi-dimensional features using a private key.

[0059] It should be noted that the multi-party secure computation protocol allows multiple participants to collaboratively complete joint computations of encrypted data without disclosing their respective original data. The encrypted computation result refers to the intermediate or final encrypted result output after joint computation of encrypted corporate debtor data, such as financial statements, tax records, and credit information, based on the multi-party secure computation protocol. The corporate debtor encrypted profile feature vector refers to a high-dimensional encrypted vector formed by integrating multi-dimensional encrypted features in a standardized format, used to characterize the comprehensive risk characteristics of corporate debtors. The multi-dimensional features of the corporate debtor encrypted profile feature vector may include tax and financial cross-features, energy consumption and operating status features, cross-institutional financial attribute features, and joint risk assessment features. The authorized result user refers to an entity legally authorized by the data provider and possessing decryption permissions, such as a bank's risk control system or a guarantee institution's decision-making platform.

[0060] It should be understood that each participating party only inputs encrypted partial data into the system, such as encrypted corporate tax records and encrypted credit records, without exposing the original plaintext. The final integrated encrypted profile feature vector remains encrypted at every intermediate stage before being decrypted by the authorized user, fundamentally avoiding the risk of data leakage. Based on the mathematical characteristics of the multi-party secure computation protocol, the computation process only operates on the encrypted data, and the decryption of the result relies on the authorized user's private key, forming a secure closed loop where data does not leave the local machine and the result is usable but invisible. Taking a financial risk control scenario as an example, the traditional model requires companies to submit a complete set of plaintext financial data to banks, posing a risk of data misuse; however, through this step, banks can only obtain the integrated encrypted feature vector, such as "encrypted value of asset-liability ratio" and "encrypted value of tax compliance level," which, after authorized decryption, yields plaintext features used for risk control modeling. This satisfies the business's need for multi-dimensional data while avoiding excessive exposure of the original data.

[0061] In another embodiment, after enabling the authorized result user to decrypt and obtain the multi-dimensional features of the corporate debtor's plaintext profile using a private key, the method further includes: The risk interpretation is performed on the plaintext profile features of the corporate debtors with the aforementioned multi-dimensional characteristics, and risk control decision suggestions are output. The risk interpretation is generated based on the correlation analysis of the multi-dimensional features, including risk level, explanation basis and business suggestions.

[0062] It should be noted that risk interpretation refers to the process of transforming data features into business-understandable risk semantic information based on multi-dimensional plaintext characteristics, through a combination of quantitative analysis and qualitative judgment. Specifically, numerical features are converted into risk meanings; for example, a debt-to-asset ratio of 0.65 implies "moderate to low debt repayment ability." The interaction of multiple features, such as "high debt-to-asset ratio" + "low tax rating," leads to "cash flow strain risk." Interpretation rules are adjusted according to the company's industry; for example, in manufacturing / internet, the weight of "R&D investment ratio" is higher for internet companies than for manufacturing companies. Correlation analysis refers to the technical process of using statistical methods or machine learning models to uncover correlations, causal relationships, or risk transmission paths between multi-dimensional features. Risk level refers to the standardized classification of the overall risk level of a corporate debtor, usually expressed as ordinal levels or score ranges. It includes two core elements: classification standards and dynamic adjustment. Classification standards refer to risk thresholds established based on industry benchmarks and historical default data; dynamic adjustment refers to real-time calibration of level boundaries based on macroeconomic data to ensure the timeliness of risk assessment.

[0063] The explanation basis refers to the traceability of risk level and risk control decisions, including the contribution of key features, rule triggering paths, and model decision logic. The contribution of key features quantifies the impact of each feature on the risk level; the rule triggering path, if using a rule engine, requires recording the specific business rules that are triggered; and the model decision logic, if using a machine learning model, explains the core basis of the model's judgment through interpretable technologies such as SHAP values ​​and LIME. Business recommendations are implementation plans tailored to different risk levels, with clear business focus. For example, for low-risk enterprises, recommendations include "simplifying the credit granting process and offering preferential interest rates"; for medium-risk enterprises, recommendations include "requiring additional collateral and shortening the post-loan inspection cycle"; and for high-risk enterprises, recommendations include "refusing credit and including them in the risk warning list."

[0064] In addition, refer to Figure 3 This invention also proposes a cross-domain profile feature extraction system for corporate debtors that combines multi-party secure computation. The system includes a cross-domain profile feature extraction device for corporate debtors that combines multi-party secure computation. The device includes a memory, a processor, and a cross-domain profile feature extraction program for corporate debtors that combines multi-party secure computation, stored in the memory and executable on the processor. The cross-domain profile feature extraction program for corporate debtors that combines multi-party secure computation is configured to implement the steps of the cross-domain profile feature extraction method for corporate debtors that combines multi-party secure computation as described above.

[0065] Other embodiments or specific implementations of the cross-domain profiling feature extraction system for corporate debtors that combines multi-party secure computation described in this invention can be referred to the above-mentioned method embodiments, and will not be repeated here.

[0066] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system 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 system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0067] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. In the unit claims listing several devices, several of these devices may be embodied by the same hardware item. The use of the terms first, second, and third, etc., does not indicate any order and can be interpreted as names.

[0068] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as a read-only memory image (ROM) / random access memory (RAM), magnetic disk, optical disk), and includes several instructions to cause a terminal user device (which may be a mobile phone, computer, server, air conditioner, or network user device, etc.) to execute the methods described in the various embodiments of the present invention.

[0069] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation, characterized in that, The method includes: Obtain the original corporate debtor data stored in the local database of each participating party, wherein each participating party includes commercial banks, tax bureaus and power grid companies, and the original corporate debtor data includes basic corporate information, financial account data, tax data and energy consumption data; Each participant is connected to the multi-party secure computing system through a secure communication channel. When the central scheduler sends a feature generation computing task instruction to each participant, the system generates the encryption key pair required for the multi-party secure computing protocol based on threshold cryptography or a trusted third party. The public key is distributed to each participant, and the private key is kept by the authorized result user. The original data of the corporate debtors of each participating party are encrypted using a public key to obtain encrypted corporate debtor data. The importance of the encrypted corporate debtor data is scored, and the encrypted corporate debtor data with an importance score greater than a preset score threshold is preferentially selected to perform secure computation using the corresponding multi-party secure computation technology combination. Based on the encrypted computation results output by the multi-party secure computation protocol, an encrypted profile feature vector of the corporate debtor containing multi-dimensional features is formed. The encrypted profile feature vector of the corporate debtor is then output to the authorized result user, so that the authorized result user can obtain the plaintext profile features of the corporate debtor with multi-dimensional features by decrypting with a private key.

2. The method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation as described in claim 1, characterized in that, The step of prioritizing the use of corresponding multi-party secure computation techniques to perform secure computation on encrypted corporate debtor data whose importance score is greater than a preset score threshold includes: The system reads encrypted corporate debtor data whose importance score is greater than a preset score threshold in blocking mode, and performs secure computation on the encrypted corporate debtor data read in blocking mode by preferentially selecting the corresponding combination of multi-party secure computation techniques.

3. The method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation as described in claim 2, characterized in that, After prioritizing the use of corresponding multi-party secure computation technology combinations to perform secure computation on encrypted corporate debtor data whose importance scores exceed a preset scoring threshold, the process further includes: Read encrypted corporate debtor data whose importance score is less than a preset score threshold using non-blocking mode; After the encrypted corporate debtor data read in the blocking mode is subjected to security computation using a combination of corresponding multi-party secure computation techniques, the encrypted corporate debtor data read in the non-blocking mode is subjected to security computation using a combination of corresponding multi-party secure computation techniques.

4. The method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation as described in claim 1, characterized in that, The step of prioritizing the use of corresponding multi-party secure computation techniques to perform secure computation on encrypted corporate debtor data whose importance score is greater than a preset score threshold includes: The computational complexity, real-time requirements, and current network conditions of encrypted corporate debtor data whose importance scores exceed a preset scoring threshold are determined. The allocation ratio of computing resources between edge devices and cloud servers is determined based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data. Based on the allocation ratio of computing resources between the edge devices and cloud servers, priority is given to selecting the corresponding combination of multi-party secure computing technologies to perform secure computing.

5. The method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation as described in claim 4, characterized in that, The process of determining the allocation ratio of computing resources between edge devices and cloud servers based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data includes: Based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data, the encrypted corporate debtor data is divided into multiple sub-task data packets; The allocation ratio of computing resources between edge devices and cloud servers is determined based on the subtask data packets.

6. The method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation as described in claim 5, characterized in that, The encrypted corporate debtor data is divided into multiple sub-task data packets based on the computational complexity, real-time requirements, and current network conditions, including: Based on the computational complexity, real-time requirements, and current network conditions of the encrypted corporate debtor data, the encrypted corporate debtor data is divided into multiple functional modules according to function or logic. Determine the dependencies and data flow relationships between the various functional modules; Multiple subtask data packets are determined based on the dependencies and data flow relationships between the functional modules.

7. The method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation as described in claim 5, characterized in that, The process of determining the allocation ratio of edge device and cloud server computing resources based on the subtask data packet includes: Based on the characteristics and resource requirements of the subtask data packets, determine the corresponding subtasks that the encrypted corporate debtor data node and its extended modules can undertake; The data calculation execution order is determined based on the corresponding sub-tasks that the encrypted corporate debtor data node and its extended modules can undertake. The allocation ratio of computing resources between edge devices and cloud servers is determined based on the execution order of the data.

8. The method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation as described in claim 1, characterized in that, The multi-dimensional corporate debtor encrypted profile feature vector includes tax and financial cross-features, energy consumption and operating status features, cross-institutional financial attribute features, and joint risk assessment features.

9. The method for extracting cross-domain profile features of corporate debtors by combining multi-party secure computation as described in claim 1, characterized in that, After enabling the authorized user to decrypt the multi-dimensional features of the corporate debtor's plaintext profile using a private key, the process further includes: The risk interpretation is performed on the plaintext profile features of the corporate debtors with the aforementioned multi-dimensional characteristics, and risk control decision suggestions are output. The risk interpretation is generated based on the correlation analysis of the multi-dimensional features, including risk level, explanation basis and business suggestions.

10. A cross-domain profiling feature extraction system for corporate debtors that combines multi-party secure computation, characterized in that, The system includes a cross-domain profile feature extraction device for corporate debtors that incorporates multi-party secure computation. The device includes a memory, a processor, and a cross-domain profile feature extraction program for corporate debtors that incorporates multi-party secure computation, stored in the memory and executable on the processor. The cross-domain profile feature extraction program for corporate debtors that incorporates multi-party secure computation is configured to implement the steps of the cross-domain profile feature extraction method for corporate debtors that incorporates multi-party secure computation as described in any one of claims 1 to 9.

Images