Multi-party fund authority supervision method and device, equipment, storage medium and product

Abstract

This application provides a method, apparatus, device, storage medium, and product for multi-party fund access supervision, relating to the field of fund management technology. The method includes: acquiring multi-party transaction data of a target transaction behavior on different business platforms; extracting transaction feature information from the multi-party transaction data, including transaction amount, transaction time, and participant identifiers; generating a risk level evaluation result corresponding to the transaction feature information through a scenario recognition rule model and a risk prediction model; and dynamically adjusting the access configuration corresponding to the target transaction behavior based on the risk level evaluation result, including the review level and identity verification requirements. This method, through the synergistic effect of dynamic access modeling and intelligent prediction, solves the shortcomings of existing technologies in terms of fixed access, delayed risk warnings, and limited scenario adaptability, achieving the goals of real-time, flexible, and secure fund supervision.

Description

Technical Field

[0001] This application relates to the field of fund management technology, and in particular to a method, apparatus, equipment, storage medium and product for monitoring the authority of multiple parties for funds. Background Technology

[0002] In the current financial and corporate fund management field, the need for collaborative fund supervision involving multiple parties is becoming increasingly complex. For example, in joint account management scenarios, multiple enterprises or departments may jointly hold and manage a fund pool, and it is necessary to ensure that the use of funds complies with preset rules; in cross-departmental collaboration scenarios, fund transfers between different departments need to be reviewed by multiple parties to avoid unauthorized misappropriation of funds; and fund flows between enterprises and their upstream and downstream enterprises need to be monitored in real time.

[0003] Existing systems typically employ a fixed permission allocation model, failing to dynamically adjust approval processes based on real-time business scenarios. This results in insufficient control over high-risk transactions, while cumbersome processes reduce efficiency in low-risk scenarios. Furthermore, traditional systems lack intelligent analysis capabilities for historical transaction data, making it difficult to identify abnormal behavior patterns in advance and proactively warn of potential risks. In multi-party collaboration scenarios, significant differences in permission rules and approval processes across different business systems lead to insufficient system compatibility and flexibility, making it difficult to adapt to complex business needs.

[0004] Therefore, the existing multi-party fund collaborative supervision adopts a fixed permission allocation model, which cannot dynamically adjust the review level according to the transaction scenario, making it difficult to achieve a balance between security and efficiency. Summary of the Invention

[0005] This application provides a method, apparatus, device, storage medium, and product for multi-party fund access control, which solves the technical problem that existing multi-party fund collaborative control methods, which adopt a fixed access control model and pre-set account access rules, cannot be dynamically adjusted according to the transaction scenario, making it difficult to achieve a balance between security and efficiency.

[0006] Firstly, this application provides a method for supervising the authority of multiple parties' funds, including:

[0007] Acquire multi-party transaction data of the target transaction behavior on different business platforms, and extract transaction feature information from the multi-party transaction data, including transaction amount, transaction time and participant identification;

[0008] By using scenario recognition rule models and risk prediction models, risk level evaluation results corresponding to transaction feature information are generated;

[0009] Based on the risk level assessment results, the permission configuration corresponding to the target transaction behavior is dynamically adjusted. The permission configuration includes the review level and identity verification requirements.

[0010] Secondly, this application provides a multi-party fund access monitoring device, comprising:

[0011] The transaction feature information extraction module is used to acquire multi-party transaction data of target transaction behavior on different business platforms and extract transaction feature information from the multi-party transaction data, including transaction amount, transaction time and participant identification.

[0012] The risk level assessment result generation module is used to generate risk level assessment results corresponding to transaction feature information through scenario identification rule model and risk prediction model;

[0013] The permission configuration adjustment module is used to dynamically adjust the permission configuration corresponding to the target transaction behavior based on the risk level assessment results. The permission configuration includes the review level and identity verification requirements.

[0014] Thirdly, this application provides an electronic device, including: a processor and a memory communicatively connected to the processor;

[0015] The memory stores the instructions that the computer executes;

[0016] The processor executes computer-executable instructions stored in memory to implement any of the methods of the first aspect.

[0017] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method of any one of the first aspects.

[0018] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method of any one of the first aspects.

[0019] The multi-party fund access control method, device, equipment, storage medium, and product provided in this application achieve real-time monitoring of fund isolation, adaptive adjustment of permissions, and proactive risk warning through the synergistic effect of dynamic permission modeling and intelligent prediction. Real-time monitoring and dynamic permission adjustment ensure that high-risk transactions are intercepted without authorization, while intelligent risk prediction identifies abnormal behavior in advance, reducing risks to fund security. In low-risk scenarios, the system automatically simplifies the review process, avoiding delays caused by manual intervention; in high-risk scenarios, multi-level review and random checks ensure security, achieving a dynamic balance between efficiency and security. Through gateway and adapter modes, the system can seamlessly interface with external systems such as bank systems and platforms, supporting multi-party data interaction, and reducing data transmission latency through edge computing nodes to ensure stability in high-concurrency scenarios. Modular design and cloud-native architecture support rapid switching between different business scenarios, and the rule customization platform allows users to customize permission rules and review processes, adapting to complex needs without code development. It solves the shortcomings of existing technologies in terms of fixed permissions, delayed risk warnings, and limited scenario adaptability, achieving the goals of real-time, flexible, and secure fund supervision. Attached Figure Description

[0020] 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.

[0021] Figure 1 A diagram illustrating the existing fund supervision system's review process;

[0022] Figure 2 A flowchart illustrating a multi-party fund access control method provided in this application embodiment;

[0023] Figure 3 A schematic diagram of a multi-party fund access monitoring device provided in this application embodiment;

[0024] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0025] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0026] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0027] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, have taken necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize or refuse.

[0028] Furthermore, the technical solution involved in this application, which involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.) and the use of artificial intelligence technology for automated decision-making, and makes decisions that have a significant impact on personal rights based on the results of automated decision-making, provides users with corresponding operation entry points for users to choose to agree to or reject the results of automated decision-making; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0029] It should be noted that the multi-party fund authority monitoring method, device, equipment, storage medium and product provided in this application can be used in the field of fund management technology, or in any field other than the field of fund management technology. The application field of the multi-party fund authority monitoring method, device, equipment, storage medium and product in this application is not limited.

[0030] The specific application scenarios of this application are applicable to financial scenarios requiring multi-party collaborative supervision. In the current financial and corporate fund management field, the need for multi-party collaborative fund supervision is becoming increasingly complex. For example, in joint account management scenarios, multiple enterprises or departments may jointly hold and manage a fund pool, requiring assurance that the use of funds complies with preset rules (e.g., a special account is only used for specific expenditures); in cross-departmental collaboration scenarios, fund transfers between different departments require multi-party review to prevent unauthorized misappropriation; in supply chain finance scenarios, the flow of funds between core enterprises and upstream and downstream enterprises needs to be monitored in real time to prevent issues such as broken cash flow chains. Furthermore, scenarios such as fund management and special fund supervision also face similar challenges.

[0031] Existing technologies pose a dual challenge to fund security and operational efficiency in financial scenarios involving multi-party collaborative supervision, such as joint account management, cross-departmental fund allocation, supply chain finance, and fund supervision. Figure 1 This is a diagram illustrating the existing fund supervision system's review process, such as... Figure 1 As shown, existing fund supervision systems mainly employ two methods: fixed permission allocation or a basic rule engine. The fixed permission allocation model involves pre-setting account permission rules (e.g., account A can only be reviewed by department managers), which cannot be dynamically adjusted according to transaction scenarios. Such systems typically rely on manual review processes, with fixed review levels and nodes, resulting in a lack of timely intervention for high-risk transactions, while low-risk scenarios suffer from inefficiency due to lengthy processes. Some systems have introduced a basic rule engine, triggering reviews based on preset conditions (e.g., transaction amount thresholds), but the rule logic is simplistic and difficult to adapt to complex business scenarios (e.g., cross-departmental joint reviews).

[0032] Therefore, existing fund supervision systems have the following shortcomings: The fixed permission allocation model of existing systems cannot dynamically adjust the review level according to business scenarios, resulting in a lack of timely control over high-risk transactions and low review efficiency in low-risk scenarios. Existing technologies lack intelligent analysis capabilities based on historical data and behavioral patterns, failing to proactively identify potential risks, leading to delayed risk warnings or high false alarm rates. Existing systems are difficult to adapt to complex business scenarios such as joint account management and cross-departmental collaboration; permission rules and review processes need to be configured separately, lacking a unified framework to support rapid switching and flexible expansion. Existing technologies may sacrifice security when simplifying the review process (e.g., automatically approving low-risk transactions), while reducing efficiency when strengthening control (e.g., multi-level review delays); a dynamic balance between the two is needed. Existing systems rely on post-event auditing for monitoring the segregation status of funds, failing to track fund flows in real time and promptly intercept illegal operations.

[0033] The multi-party fund access control method, device, equipment, storage medium, and product provided in this application utilize distributed data acquisition and streaming processing technologies to acquire multi-party transaction data in real time; dynamically evaluate transaction scenarios and risk levels through a multi-level rule engine and machine learning model; and adapt to access control and approval processes in different business scenarios by combining modular design and cloud-native architecture. Through the synergy of dynamic modeling and intelligent prediction, it resolves the security and efficiency contradictions under existing fixed access control models and achieves flexible adaptation to complex business scenarios, aiming to solve the aforementioned technical problems of existing technologies.

[0034] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0035] Figure 2 This is a flowchart illustrating a multi-party fund access control method provided in an embodiment of this application, as shown below. Figure 2 As shown, the method includes:

[0036] S201. Obtain multi-party transaction data of the target transaction behavior on different business platforms, and extract transaction feature information from the multi-party transaction data.

[0037] In this embodiment of the application, the transaction feature information includes the transaction amount, transaction time, and participant identifiers.

[0038] In one example, transaction feature information is a multidimensional data set describing transaction behavior, used to characterize the contextual features of the transaction.

[0039] S202. By using the scenario recognition rule model and risk prediction model, generate risk level evaluation results corresponding to transaction feature information.

[0040] In one example, the risk level assessment result is a quantitative risk level output after analyzing the transaction characteristics through a model, which is used to characterize the potential risk level of the target transaction behavior.

[0041] S203. Based on the risk level assessment results, dynamically adjust the permission configuration corresponding to the target transaction behavior.

[0042] In this embodiment of the application, the permission configuration includes the audit level and the authentication requirements.

[0043] In one example, the permission configuration is a set of permission parameters that are dynamically set based on the risk level score, including the review level (such as basic review or high-level review) and identity verification requirements (such as SMS verification code or biometric authentication).

[0044] In one implementation scenario, a multi-party fund collaborative supervision system based on dynamic permission modeling and intelligent prediction is used to achieve multi-party fund collaborative supervision. The multi-party fund collaborative supervision system may include: a fund isolation real-time monitoring module, a dynamic permission adjustment module, an intelligent risk prediction module, an operation efficiency optimization module, and a multi-scenario adaptation module.

[0045] For example, the real-time fund segregation monitoring module is specifically used to: connect to banking systems, payment platforms, and internal accounting systems via interfaces to obtain real-time fund transaction data for target transactions, including transaction amount, time, and account information. It presets fund segregation rules based on business scenarios, such as specific accounts being used only for specific purposes and fund transfers requiring multi-party signature verification. It continuously monitors fund flows, and immediately triggers alarms and records anomaly logs upon detecting any violation of preset fund segregation rules (such as unauthorized fund transfers). It provides managers with a visual dashboard of fund flows, displaying the segregation status and fund distribution of each account.

[0046] The dynamic permission adjustment module is specifically used to: identify the current business scenario (such as daily small payments or large-sum payments) by analyzing information such as transaction type, amount, and participants. Combined with the output of the intelligent risk prediction module, it determines the risk level of the current transaction, classifying it into low, medium, and high levels. Permissions are automatically adjusted according to the risk level; for example, low-risk transactions require only basic review, medium-risk transactions require confirmation from middle-level managers, and high-risk transactions are temporarily escalated to the level of senior decision-makers. The system automatically executes the permission adjustment and generates adjustment records for subsequent auditing, while simultaneously notifying relevant personnel of the new review requirements. Based on actual operational feedback, the permission adjustment rules are continuously optimized to ensure that the adjustment logic aligns with actual business needs.

[0047] The intelligent risk prediction module is specifically used to: collect historical transaction data, user operation records, and external risk intelligence to form a multi-dimensional dataset. It employs machine learning algorithms (such as random forests or neural networks, which are not limited in this application) to train a risk prediction model, identifying characteristic patterns of high-risk transactions, such as abnormal transaction times and frequent large-amount transfers. The system analyzes current transaction data in real time and, combined with the model, outputs risk probabilities and levels. Based on the prediction results, the system automatically generates permission adjustment suggestions, such as suggesting adding review levels or requiring additional identity verification. The dataset is regularly updated to optimize model parameters and improve prediction accuracy.

[0048] The operational efficiency optimization module is specifically used to categorize business scenarios into low-risk and non-low-risk types based on dimensions such as transaction amount, frequency, and participants. For low-risk scenarios, a simplified approval process is designed; for example, small-amount daily expense reimbursements can be automatically approved without manual intervention. For non-low-risk scenarios, the module's execution is adjusted according to dynamic permissions. The module records the time and results of each approval, analyzes process bottlenecks, and proposes optimization suggestions. It also provides operation guidance and intelligent reminders to help users quickly complete approvals or understand current permission requirements.

[0049] The multi-scenario adaptation module is specifically designed to establish permission allocation and fund supervision models for scenarios such as joint account management, cross-departmental collaboration, and third-party escrow. It allows users to customize permission levels, approval processes, and fund segregation rules according to specific business needs. It ensures seamless integration with different business systems (such as enterprise resource planning systems and payment platforms) and supports multi-party data interaction. It supports rapid switching of permission and monitoring modes between different scenarios to adapt to business changes. It generates detailed operation logs and permission adjustment records for each scenario to meet compliance requirements.

[0050] In another implementation scenario, multi-party fund access control methods may include:

[0051] The system connects in real-time to bank systems and payment platform interfaces via a distributed data acquisition gateway to obtain multi-party transaction data (including transaction amount, time, account information, participants, etc.). This data is then analyzed in real-time using a streaming processing framework, combined with a multi-level fund segregation rule engine (such as account usage restrictions and multi-party signature verification) to determine if segregation rules are violated. Real-time fund flow status (compliant / abnormal) and abnormal transaction logs are output. For example, in a joint account management scenario, if the system detects an attempt to transfer funds from a dedicated account to a non-designated account, it immediately triggers an alarm, suspends the transaction, and notifies relevant personnel for review. The distributed data acquisition gateway is a lightweight data acquisition module deployed on the transaction terminal, supporting real-time access to multi-source data (bank systems, payment platforms). The streaming processing framework is used to process continuous data streams in real-time and perform complex event analysis. The multi-level fund segregation rule engine is a rule base configured based on dimensions such as account type, purpose, and amount range to determine whether fund transfers are compliant. It enables real-time tracking of fund flows and interception of illegal operations. Through distributed data collection and streaming processing technology, it achieves millisecond-level response and, combined with a rule engine, ensures that the use of funds complies with preset conditions, ultimately improving fund security and regulatory efficiency.

[0052] The system takes transaction scenario characteristics (such as transaction amount, time, and participants) and risk level scores as input, and outputs permission adjustment instructions through a scenario recognition rule base and risk prediction model (such as machine learning-based classification algorithms). It then uses a state machine framework to dynamically switch permission levels (e.g., low-risk scenarios require only basic-level review, while high-risk scenarios require intervention from senior decision-makers), outputting dynamic permission configurations (including but not limited to review levels and identity verification requirements). For example, in cross-departmental collaborative fund disbursement, if the system detects a large transaction involving multiple external accounts and rates it as "high," it automatically elevates the review authority from the department manager to the CFO. The transaction can only proceed after the CFO's confirmation. The scenario recognition rule base consists of preset business scenario classification rules (such as transaction amount, time, and participant combinations) used to match risk levels. The risk prediction model is a machine learning model (such as random forests or neural networks) trained on historical data used to evaluate transaction risk levels. The state machine framework provides the logical structure for switching permission states based on risk levels, ensuring clear and controllable adjustment logic.

[0053] In another implementation scenario, lightweight edge computing nodes can be deployed near the transaction terminal. Combined with a streaming processing framework, data processing logic can be optimized to achieve millisecond-level monitoring and anomaly interception of fund flows. Specifically, this can include: deploying edge computing devices at transaction terminals (such as ATMs and enterprise payment systems) to collect transaction data in real time and perform preliminary filtering (such as amount threshold judgment). The granularity of data processing can be dynamically adjusted through the windowing mechanism of the streaming processing framework, combined with in-memory database caching of high-frequency transaction characteristics to reduce redundant calculations. Commonly used fund isolation rules (such as account usage restrictions) can be preloaded on edge nodes, while only complex rules (such as multi-signature verification) are uploaded to the cloud for processing, reducing transmission latency.

[0054] By reducing data transmission time to the cloud through edge computing and combining it with the dynamic windowing mechanism of streaming processing, the system can complete anomaly detection and interception within a very short time (e.g., within 100ms) after a transaction occurs. In a unified account management scenario, when funds attempt to be transferred from a dedicated account to a non-designated account, the edge node can directly trigger a local alarm and suspend the transaction without waiting for a response from the cloud, thereby significantly improving real-time performance and reducing system load in high-concurrency scenarios.

[0055] In another implementation scenario, a multi-model fusion architecture (such as integrating random forests, neural networks, and time series models) can be introduced, combined with a dynamic weight adjustment mechanism, to improve the accuracy and adaptability of risk prediction. Specifically, this can include: training multiple risk prediction models separately; for example, using random forests to classify transaction types, LSTM (Long Short-Term Memory) networks to analyze time series features, and graph neural networks to identify account correlations; and then outputting a comprehensive risk score through model fusion. The weight ratios of each model are dynamically adjusted according to the characteristics of the business scenario (such as supply chain finance and charitable fund management). For example, in supply chain finance, time series models have higher weights (focusing on transaction frequency), while in charitable fund management, graph neural networks have higher weights (focusing on account correlations). Incremental learning techniques are used to update model parameters in real time to adapt to emerging risk patterns (such as new methods), preventing the model from becoming outdated.

[0056] By fusing multiple models, the system covers risk characteristics across different dimensions (such as transaction frequency, account correlation, and abnormal amounts), and dynamic weight adjustments ensure that the model output better meets the needs of specific business scenarios. The system can identify abnormal fund flows between upstream and downstream enterprises in advance and enhance the sensitivity of the time series model through dynamic weights, thereby triggering access control upgrades before risk events occur and reducing potential losses.

[0057] In another implementation scenario, a node-edge model of fund flows can be constructed using a graph database. Graph traversal algorithms (such as depth-first search) can then be used to identify complex and abnormal paths (such as fund transfers through intermediary accounts). Specifically, this can involve modeling entities such as accounts, transaction records, and participants as graph nodes, and transaction relationships as edges, storing all fund flow paths. Graph traversal algorithms are then used to detect abnormal path characteristics (such as funds being transferred to abnormal accounts through multiple intermediary accounts). By combining timestamps and amount information, the temporal continuity and amount distribution patterns of fund flows can be analyzed to identify concealed irregularities.

[0058] Graph databases can visually represent the complex paths of fund flows, and graph traversal algorithms can quickly detect hidden anomalies. In a joint account management scenario, the system detects that a sum of money flows into an abnormal account through three intermediary accounts. Although the amount of each individual transaction does not exceed the threshold, the path characteristics trigger an anomaly alarm, thereby effectively intercepting concealed violations.

[0059] In another implementation scenario, caching mechanisms and asynchronous processing technologies can be introduced into the dynamic permission adjustment module to reduce response latency and ensure business process continuity. Specifically, this can include: preloading commonly used permission adjustment rules (such as review levels for low-risk scenarios) into a memory cache to avoid frequent database queries; decoupling the execution of permission adjustment instructions from the transaction process and processing permission adjustment tasks asynchronously through message queues to ensure that the transaction process is not blocked; and dynamically adjusting the priority of message queues based on risk levels (e.g., prioritizing high-risk transactions) to ensure the real-time nature of critical permission adjustments.

[0060] Caching technology reduces permission rule query time, and asynchronous processing prevents transaction delays caused by permission adjustments. The system can quickly adjust the approval level (e.g., from department manager to CFO) through asynchronous processing without affecting other transaction processes (e.g., freezing funds), thereby improving operational efficiency while ensuring security.

[0061] In another implementation scenario, federated learning technology can be used to enable multiple participants (such as banks, enterprises, and regulatory agencies) to collaboratively train a risk prediction model without sharing raw data, thereby improving the comprehensiveness and privacy of predictions. Specifically, this can include: synchronizing model parameters among participants using encrypted communication protocols (such as isomorphic encryption), rather than raw transaction data, to ensure data privacy. Each participant trains its own risk prediction model locally (such as a random forest model based on its own transaction data), and the global model is generated through parameter aggregation. The global model parameters are periodically synchronized and updated to adapt to differences in risk patterns among different institutions (e.g., the anomalous characteristics of a bank differ from the anomalous transaction patterns of a company).

[0062] Federated learning enables participating parties to share risk prediction capabilities while protecting data privacy, thereby improving the model's ability to identify cross-institutional risks. Core enterprises and upstream and downstream enterprises can collaboratively train the model through federated learning, which can identify abnormal cross-institutional fund flows in advance (such as a supplier suddenly making frequent transfers to related accounts), thereby triggering access control escalation before the risk spreads and reducing systemic risk.

[0063] The multi-party fund access control method provided in this embodiment achieves real-time monitoring of fund isolation, adaptive adjustment of permissions, and proactive risk warning through the synergistic effect of dynamic permission modeling and intelligent prediction. Real-time monitoring and dynamic permission adjustment ensure that high-risk transactions are intercepted without authorization, while intelligent risk prediction identifies abnormal behavior in advance, reducing risks to fund security. In low-risk scenarios, the system automatically simplifies the review process, avoiding delays caused by manual intervention; in high-risk scenarios, multi-level review and random checks ensure security, achieving a dynamic balance between efficiency and security. Through gateway and adapter modes, the system can seamlessly interface with external systems such as bank systems and platforms, supporting multi-party data interaction, and reducing data transmission latency through edge computing nodes to ensure stability in high-concurrency scenarios. Modular design and cloud-native architecture support rapid switching between different business scenarios, and the rule customization platform allows users to customize permission rules and review processes, adapting to complex needs without code development. This addresses the shortcomings of existing technologies in terms of fixed permissions, delayed risk warnings, and limited scenario adaptability, achieving the goals of real-time, flexible, and secure fund supervision.

[0064] Optionally, risk level evaluation results corresponding to transaction feature information are generated through scenario recognition rule model and risk prediction model, including: matching transaction feature information with preset business scenarios through scenario recognition model to obtain scenario classification results of target transaction behavior; analyzing historical transaction data and behavior patterns through risk prediction model to obtain risk probability distribution of target transaction behavior; and weightedly integrating scenario classification results and risk probability distribution to generate risk level evaluation results.

[0065] In one example, a scenario recognition model maps transaction characteristics (such as amount, time, and participants) to specific business scenarios (such as daily payments or cross-departmental transfers). A risk prediction model analyzes historical transaction data to calculate risk probabilities. A weighted fusion model combines the scenario classification results with the risk probabilities to generate a risk level assessment result. For example, when transaction characteristics match the "large cross-departmental transfer" scenario and the risk prediction model outputs a high risk probability, the weighted fusion model outputs a "high" risk level score for subsequent permission adjustments. The scenario recognition model is a classification algorithm based on preset rules or a machine learning model, used to map transaction characteristics to specific business scenarios. The risk prediction model is a machine learning model (such as a random forest or neural network) trained on historical transaction data, used to evaluate the potential risk probability of a transaction. The weighted fusion model, combining scenario classification results with risk probabilities using a fusion algorithm (such as linear weighting or decision trees), generates the final risk level assessment result.

[0066] By leveraging the synergy of a scene recognition model and a risk prediction model, accurate risk level scoring is achieved. The scene recognition model matches transaction characteristics with specific business scenarios using pre-defined rules or machine learning classification algorithms, ensuring the scenario adaptability of the risk assessment. The risk prediction model calculates risk probabilities based on historical data, enhancing the dynamics and adaptability of the scoring. The weighted fusion model combines scene classification results and risk probabilities through algorithms to generate the final risk level score. This ensures that the score reflects both the characteristics of the transaction scenario and historical risk patterns, significantly improving the accuracy of the risk level scoring and its adaptability to business scenarios. This provides a more reliable basis for the dynamic adjustment of subsequent permission configurations.

[0067] Optionally, the analysis of historical transaction data and behavioral patterns through a risk prediction model specifically includes at least one of the following: using an anomaly detection model based on time series analysis to detect anomalies in the target transaction behavior based on transaction amount and transaction time; using an account correlation analysis model based on graph neural networks to construct an account-transaction relationship graph of the target transaction behavior and identify the correlation between accounts in the target transaction behavior; and using a multi-feature classification model based on random forests to componentize the transaction risk level corresponding to the target transaction behavior based on transaction feature information.

[0068] In one example, three optional risk prediction models (time series analysis, graph neural networks, and random forests) cover different dimensions of risk characteristics. For instance, the time series analysis model detects abnormal fluctuations in transaction time and amount, the graph neural network model identifies hidden correlations between accounts, and the random forest model evaluates risk probability through multi-feature classification. These three models can be used independently or in combination to ensure comprehensive risk prediction. Specifically, the time series analysis model identifies abnormal fluctuations by analyzing the continuity of transaction time and amount. The graph neural network model identifies hidden account correlations by constructing an account-transaction relationship graph. The random forest classification model is an ensemble learning model based on multi-dimensional features (such as transaction amount, participants, and time) used to classify transaction risk levels.

[0069] In another example, a weighted fusion model dynamically adjusts the weight ratio between scenario classification results and risk probabilities based on the business scenario. The fusion weights of scenario classification results and risk probabilities are adjusted according to the differences in characteristics of business scenarios (such as supply chain finance and fund management). By dynamically adjusting the weight ratios, the weighted fusion model adapts to the risk characteristics of different business scenarios. For example, in supply chain finance, time series models have higher weights, and the system focuses more on transaction frequency; in fund management, graph neural network models have higher weights, and the system focuses more on account correlation. This dynamic adjustment of weight ratios allows the weighted fusion model to adapt to the risk characteristics of different business scenarios.

[0070] By introducing multiple types of risk prediction models, the comprehensiveness and adaptability of risk identification are improved. Time series analysis models cover abnormal fluctuations in transaction time and amount, graph neural network models identify hidden account correlations, and random forest models enhance the accuracy of risk probability through multi-feature classification. The optional combination of these three models allows the system to adapt to the risk characteristics of different business scenarios (such as supply chain finance and charitable fund management), significantly improving the robustness and scenario adaptability of risk prediction.

[0071] Optionally, acquire multi-party transaction data of the target transaction behavior on different business platforms and extract transaction feature information from the multi-party transaction data, including: deploying edge computing nodes on each business platform, collecting multi-party transaction data of the target transaction behavior on different business platforms in real time and filtering it to obtain filtered transaction data; and performing real-time analysis on the filtered transaction data through a streaming processing framework to extract transaction feature information from the multi-party transaction data.

[0072] In one example, raw transaction data (such as amount and time) is collected in real time at the trading terminal via edge computing nodes and preliminarily filtered (such as by determining amount thresholds) to reduce the amount of data transmitted to the cloud. The streaming processing framework then performs real-time analysis on the filtered data (such as extracting transaction time and participant identifiers) to generate transaction feature information, which serves as input for subsequent risk assessment. Edge computing nodes are lightweight computing devices deployed at the trading terminal for real-time collection and preliminary processing of transaction data. The streaming processing framework is a computing framework for processing continuous data streams in real time, supporting low-latency analysis.

[0073] In another example, a federated learning framework enables multiple participants (such as banks and enterprises) to train risk prediction models locally and synchronize model parameters via encrypted communication protocols to generate a global model. For instance, banks and enterprises train models based on their own transaction data, and the federated learning framework aggregates the model parameters, improving the cross-institutional adaptability of risk prediction. The federated learning framework synchronizes model parameters among multiple participants using encrypted communication protocols (such as isomorphic encryption), rather than the raw data. This enables collaborative training of cross-institutional risk prediction models, improving the comprehensiveness and adaptability of the models while protecting data privacy. Participants do not need to share raw transaction data, only synchronizing model parameters, ensuring data privacy; the global model adapts to the risk patterns of different institutions through parameter aggregation, significantly improving the accuracy and scenario adaptability of cross-institutional risk identification.

[0074] By leveraging the synergy of edge computing and streaming processing, the efficiency and real-time performance of acquiring transaction feature information are significantly improved. Edge computing nodes reduce data transmission latency, while the streaming processing framework ensures low latency feature extraction, enabling the system to complete feature information extraction within a very short time after a transaction occurs, providing real-time input for subsequent risk assessment and permission adjustments.

[0075] Optionally, the method also includes: loading preset account usage restriction rules on edge computing nodes and performing compliance judgment on transaction data in the streaming processing framework.

[0076] In one example, commonly used fund segregation rules (such as account usage restriction rules) are preloaded on edge computing nodes, enabling preliminary compliance checks on transaction data within the streaming processing framework. For instance, if an attempt to transfer funds from a dedicated account to a non-designated account is detected, the edge computing node directly triggers a local alarm and suspends the transaction. The account usage restriction rules are preset rules (such as "dedicated accounts are only for specified expenditures").

[0077] By pre-loading rules and conducting preliminary compliance checks, the real-time performance and interception efficiency of fund segregation monitoring are significantly improved. Edge computing nodes can complete compliance checks as transactions occur without waiting for cloud responses, ensuring immediate interception of violations and reducing system load in high-concurrency scenarios.

[0078] Optionally, based on the risk level assessment results, the permission configuration corresponding to the target transaction behavior can be dynamically adjusted, including: matching a preset permission template based on the risk level assessment results, obtaining the review and verification level and identity verification requirements corresponding to the target transaction behavior; dynamically switching the permission configuration corresponding to the target transaction behavior through a state machine framework, and recording the permission adjustment log corresponding to the target transaction behavior.

[0079] In one example, the risk level assessment results are matched with a permission template, and the corresponding permission configuration (such as review level and identity verification requirements) is output. The state machine framework dynamically switches permission rules according to the configuration and records permission adjustment logs. For example, when the risk level score is "high," the state machine framework switches to the "multi-level review + biometric verification" permission configuration and records the adjustment logs for subsequent auditing. The preset permission template includes a set of preset permission configuration rules (such as "low-risk scenario → automatic review" and "high-risk scenario → multi-level review + biometric verification"). The state machine framework is a permission management architecture based on state transition logic, used for dynamically switching permission configurations.

[0080] By combining permission templates and a state machine framework, dynamic and controllable permission configuration is achieved. Permission templates ensure standardized permission adjustment rules, while the state machine framework guarantees the stability and traceability of the adjustment process through state transition logic. Ultimately, this step significantly improves operational efficiency by simplifying permissions in low-risk scenarios while ensuring fund security, achieving a dynamic balance between security and efficiency.

[0081] Figure 3 A schematic diagram of a multi-party fund access monitoring device provided in this application embodiment is shown below. Figure 3As shown, the multi-party fund access control device 30 provided in this embodiment includes:

[0082] The transaction feature information extraction module 301 is used to acquire multi-party transaction data of the target transaction behavior on different business platforms and extract the transaction feature information of the multi-party transaction data, including transaction amount, transaction time and participant identification.

[0083] The risk level assessment result generation module 302 is used to generate risk level assessment results corresponding to transaction feature information through scenario identification rule model and risk prediction model;

[0084] The permission configuration adjustment module 303 is used to dynamically adjust the permission configuration corresponding to the target transaction behavior based on the risk level assessment results. The permission configuration includes the review level and identity verification requirements.

[0085] In one possible implementation, the risk level assessment result generation module 302 is specifically used to: match transaction feature information with preset business scenarios through a scenario recognition model to obtain the scenario classification result of the target transaction behavior; analyze historical transaction data and behavior patterns through a risk prediction model to obtain the risk probability distribution of the target transaction behavior; and weight and fuse the scenario classification result and the risk probability distribution to generate the risk level assessment result.

[0086] In one possible implementation, the risk level assessment result generation module 302 is further specifically used to: detect anomalies in the target transaction behavior using an anomaly detection model based on time series analysis, based on the transaction amount and transaction time; construct an account-transaction relationship graph of the target transaction behavior using an account correlation analysis model based on graph neural networks, and identify the correlation between accounts in the target transaction behavior; and use a multi-feature classification model based on random forests to componentize the transaction risk level corresponding to the target transaction behavior based on transaction feature information.

[0087] In one possible implementation, the transaction feature information extraction module 301 is specifically used to: deploy edge computing nodes on each business platform, collect multi-party transaction data of target transaction behavior on different business platforms in real time and filter it to obtain filtered transaction data; and perform real-time analysis on the filtered transaction data through a streaming processing framework to extract transaction feature information of the multi-party transaction data.

[0088] In one possible implementation, the multi-party fund access control device is also specifically used to: load preset account usage restriction rules on edge computing nodes and perform compliance judgment on transaction data in the streaming processing framework.

[0089] In one possible implementation, the permission configuration adjustment module 303 is specifically used to: match a preset permission template based on the risk level assessment result, obtain the review and verification level and identity verification requirements corresponding to the target transaction behavior; dynamically switch the permission configuration corresponding to the target transaction behavior through a state machine framework, and record the permission adjustment log corresponding to the target transaction behavior.

[0090] The multi-party fund access control device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0091] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 40 may include a memory 401 and a processor 402. Optionally, the electronic device may also include a transceiver 403, wherein the memory 401 and the processor 402 communicate with each other; for example, the memory 401, the processor 402 and the transceiver 403 may communicate via a communication bus 404, the memory 401 is used to store a computer program, and the processor 402 executes the computer program to implement the method of the above embodiments.

[0092] Optionally, the processor mentioned above can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps in the method embodiments disclosed in this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0093] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the methods in any of the above method embodiments.

[0094] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the methods in any of the above method embodiments.

[0095] All or part of the steps in the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable memory. When the program is executed, it performs the steps of the above method embodiments; and the aforementioned memory (storage medium) includes: read-only memory (ROM), RAM, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof.

[0096] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processing unit of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processing unit of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0099] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

[0100] In this application, the term "comprising" and its variations can refer to non-limiting inclusion; the term "or" and its variations can refer to "and / or". The terms "first", "second", etc., in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. In this application, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0101] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0102] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0103] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0104] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or in the form of software program modules.

[0105] When integrated units / modules are implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor can be any suitable hardware processor, such as a CPU, GPU, FPGA, DSP, and ASIC, etc. Unless otherwise specified, the storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.

[0106] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0107] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0108] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0109] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for supervising the authority of multiple parties' funds, characterized in that, The method includes: Acquire multi-party transaction data of the target transaction behavior on different business platforms, and extract transaction feature information from the multi-party transaction data, including transaction amount, transaction time and participant identification; By using a scenario recognition rule model and a risk prediction model, a risk level evaluation result corresponding to the transaction feature information is generated; Based on the risk level assessment results, the permission configuration corresponding to the target transaction behavior is dynamically adjusted. The permission configuration includes the review level and identity verification requirements.

2. The method according to claim 1, characterized in that, The step of generating a risk level evaluation result corresponding to the transaction feature information through a scenario recognition rule model and a risk prediction model includes: By using the scene recognition model, the transaction feature information is matched with a preset business scenario to obtain the scene classification result of the target transaction behavior; By analyzing historical transaction data and behavioral patterns using the aforementioned risk prediction model, the risk probability distribution of the target transaction behavior can be obtained. The scenario classification results are weighted and fused with the risk probability distribution to generate the risk level evaluation results.

3. The method according to claim 2, characterized in that, The analysis of historical transaction data and behavioral patterns through the risk prediction model specifically includes at least one of the following: Based on the transaction amount and the transaction time, an anomaly detection model based on time series analysis is used to detect anomalies in the target transaction behavior. Using an account association analysis model based on graph neural networks, an account-transaction relationship graph of the target transaction behavior is constructed to identify the association relationships between accounts in the target transaction behavior; Using a multi-feature classification model based on random forest, the transaction risk level corresponding to the target transaction behavior is componentized according to the transaction feature information.

4. The method according to claim 1, characterized in that, The process of acquiring multi-party transaction data of the target transaction behavior on different business platforms and extracting transaction feature information from the multi-party transaction data includes: Edge computing nodes are deployed on each of the aforementioned business platforms to collect and filter multi-party transaction data of the target transaction behavior on different business platforms in real time, thereby obtaining filtered transaction data. The filtered transaction data is analyzed in real time using a streaming processing framework to extract transaction feature information from the multi-party transaction data.

5. The method according to claim 4, characterized in that, The method further includes: Preset account usage restriction rules are loaded on the edge computing node, and the transaction data is judged for compliance in the streaming processing framework.

6. The method according to claim 1, characterized in that, The step of dynamically adjusting the permission configuration corresponding to the target transaction behavior based on the risk level assessment result includes: Based on the risk level assessment results, a preset permission template is matched to obtain the review and verification level and identity verification requirements corresponding to the target transaction behavior; The permission configuration corresponding to the target transaction behavior is dynamically switched through a state machine framework, and the permission adjustment log corresponding to the target transaction behavior is recorded.

7. A multi-party fund access control monitoring device, characterized in that, The device includes: The transaction feature information extraction module is used to acquire multi-party transaction data of target transaction behavior on different business platforms, and extract transaction feature information from the multi-party transaction data, including transaction amount, transaction time and participant identification. The risk level assessment result generation module is used to generate the risk level assessment result corresponding to the transaction feature information through the scenario identification rule model and the risk prediction model. The permission configuration adjustment module is used to dynamically adjust the permission configuration corresponding to the target transaction behavior based on the risk level evaluation results. The permission configuration includes the review level and identity verification requirements.

8. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

Images