Method and system for multi-end data privacy collaborative processing based on central networking node

Abstract

The application provides a multi-terminal data privacy collaborative processing method and system based on a central networking node, relates to the technical field of data processing, and comprises the following steps: deploying a plurality of heterogeneous privacy computing nodes designed by different privacy computing protocols, network communication protocols and upper-layer applications; setting a central coordination node for the demand of interconnection and intercommunication of the heterogeneous privacy computing nodes, wherein the central coordination node has protocol compatibility; enabling each heterogeneous privacy computing node to be connected with the central node once through the protocol compatibility of the central node, and establishing indirect interconnection and intercommunication channels between all the privacy computing nodes through the central node. The application realizes the collaborative processing of calculation localization and management centralization.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a multi-terminal data privacy collaborative processing method and system based on a central network node. Background Technology

[0002] Against the backdrop of the rapid development of the digital economy, data from multiple fields is generally scattered across different institutions or terminals, forming data silos. Meanwhile, the improvement of privacy protection regulations has imposed strict constraints on data sharing and utilization. Although privacy computing technology has made it possible for data to be usable but not visible, the heterogeneity of privacy computing protocols, network communication protocols, and upper-layer application designs often makes it difficult for multiple participating nodes to achieve efficient collaboration.

[0003] For example, conducting joint financial risk modeling can avoid the leakage of original customer financial information and integrate data from multiple parties to improve the accuracy of risk identification. However, existing technical solutions lack core nodes with unified protocol compatibility and centralized scheduling capabilities, which requires each financial institution's privacy computing nodes to be adapted to each other's protocols. This not only increases the complexity of networking and the cost of connection, but also leads to problems such as wasted computing resources and chaotic task execution timing due to the lack of targeted task allocation and global monitoring mechanisms. It is difficult to fully release the collaborative value of multi-terminal data while ensuring data privacy. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a multi-terminal data privacy collaborative processing method and system based on a central networking node, so as to realize the collaborative processing of localized computing and centralized management.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] Firstly, a multi-terminal data privacy collaborative processing method based on a central network node, the method comprising:

[0007] Deploy multiple heterogeneous privacy computing nodes with different privacy computing protocols, network communication protocols, and upper-layer application designs;

[0008] To address the need for interconnection and interoperability among heterogeneous privacy computing nodes, a central coordination node is set up, which has protocol compatibility capabilities.

[0009] By leveraging the protocol compatibility of the central node, each heterogeneous privacy computing node can establish a network connection with the central node, thereby creating an indirect interconnection channel between all privacy computing nodes.

[0010] Based on the established interconnection channels, the central node collects source data from each node and transforms it into a multi-dimensional data point set. By performing cluster boundary analysis on the multi-dimensional data point set, the cluster boundary information of the data point set is obtained. Based on the cluster boundary information, a reference plane is constructed, the base region is defined, and interrelated sub-regions are divided according to the data distribution and cluster boundary information to form the sub-region division result.

[0011] Based on the sub-region division results, multi-dimensional data points are mapped to corresponding sub-regions. Task configuration parameters are obtained according to the aggregation characteristics of data points in each sub-region. The computation task logic of each privacy computing node is adapted and adjusted based on the task configuration parameters. The central coordination node obtains the corresponding coordination instructions and task payloads based on the adjusted task logic.

[0012] By coordinating the command flow and task payload flow for unified scheduling and distributing routes to the corresponding privacy computing nodes, each node is scheduled to execute computing tasks locally, achieving collaborative processing of computing localization and centralized management.

[0013] Furthermore, multiple heterogeneous privacy computing nodes with different privacy computing protocols, network communication protocols, and upper-layer application designs are deployed, including:

[0014] Heterogeneous characteristics identification and analysis are performed on each privacy computing node to be connected, and the analysis results of differences in privacy computing protocol type, network communication protocol type, and upper-layer application design architecture are obtained.

[0015] Based on the heterogeneity characteristic identification and analysis results, the privacy computing protocol compatibility capability is configured to support the conversion and processing of multiple privacy computing protocols such as RSA+Hash, Chinese cryptographic +Hash, ECDH, OPRF and VOLE.

[0016] Based on the configuration results of the privacy computing protocol compatibility, network communication protocol adaptation capability is established, enabling the network communication protocol adaptation capability to handle protocol conversion of one-way network communication request-pull mode, long connection mode and gRPC protocol, thus obtaining the establishment result of the network communication protocol adaptation capability.

[0017] Based on the results of establishing network communication protocol adaptation capabilities, application interface standardization capabilities are generated. This standardization capability unifies the application interface specifications of different heterogeneous privacy computing platforms, realizes the standardized mapping of heterogeneous application design styles, and yields the results of generating application interface standardization capabilities.

[0018] Based on the generated application interface standardization capabilities, the privacy computing protocol compatibility, network communication protocol adaptation, and application interface standardization capabilities are integrated to complete the unified deployment of multiple heterogeneous privacy computing nodes.

[0019] Furthermore, to address the need for interconnectivity among heterogeneous privacy computing nodes, a central coordination node is established. This central coordination node possesses protocol compatibility capabilities, including:

[0020] Based on the networking requirements for interconnection of heterogeneous privacy computing nodes, a basic architecture for a central coordination node is constructed.

[0021] Through the infrastructure of the central coordination node, a network protocol compatibility layer is configured to recognize and convert communication data of one-way network communication request-pull mode, long connection mode and gRPC protocol, and obtain the configuration result of the network protocol compatibility layer;

[0022] Based on the configuration results of the network protocol compatibility layer, a privacy computing protocol conversion capability is constructed to enable the real-time conversion and adaptation of multiple privacy computing protocols such as RSA+Hash, Chinese cryptographic + Hash, ECDH, OPRF and VOLE, thus obtaining the construction result of the privacy computing protocol conversion capability.

[0023] Based on the construction results of privacy computing protocol conversion capabilities, an application layer interface standardization mechanism is established. This mechanism maps the application interface specifications of different heterogeneous privacy computing platforms into a standard interface format that can be recognized by the central coordination node, thus obtaining the establishment result of the application layer interface standardization mechanism.

[0024] By establishing the application layer interface standardization mechanism, the network protocol compatibility layer, privacy computing protocol conversion capabilities, and application layer interface standardization mechanism are integrated to form a complete protocol compatibility capability for the central coordination node.

[0025] Furthermore, leveraging the protocol compatibility of the central node, each heterogeneous privacy computing node establishes a network connection with the central node, thereby creating an indirect interconnection channel between all privacy computing nodes. This includes:

[0026] By leveraging the complete protocol compatibility of the central coordination node, the network services of the central node are initialized, providing a unified access endpoint for heterogeneous privacy computing nodes and obtaining the network service initialization status.

[0027] Based on the initialization status of the network service, receive network connection requests from each heterogeneous privacy computing node, perform network protocol identification and protocol conversion processing on each network connection request, and obtain the converted standardized network request.

[0028] Based on the converted standardized networking request, a single networking connection is established between each heterogeneous privacy computing node and the central node, resulting in a node-central connection mapping table.

[0029] Based on the node and central connection mapping table, a global routing table for the central node is constructed, which enables the global routing table to forward data flows from any source node to the corresponding target node through the central node according to the target node identifier, thus obtaining the routing and forwarding capability of the central node;

[0030] By leveraging the routing and forwarding capabilities of the central node, the transparent forwarding mechanism of the central node's data flow is activated, enabling the central node to only perform data routing and distribution without participating in actual computation, thus establishing an indirect interconnection channel between all privacy computing nodes.

[0031] Furthermore, based on the established interconnection channels, the central node collects source data from each node and transforms it into a multi-dimensional data point set. Cluster boundary analysis is performed on the multi-dimensional data point set to obtain cluster boundary information. A reference plane is constructed based on this cluster boundary information, a base region is defined, and interrelated sub-regions are divided according to data distribution and cluster boundary information, resulting in sub-region division results, including:

[0032] Source data feature information is collected from various privacy computing nodes through interconnection channels and transformed into a standardized set of multi-dimensional data points.

[0033] Based on the standardized representation of a multidimensional data point set, a point clustering boundary extraction algorithm is executed to perform density distribution analysis and boundary identification on the multidimensional data points, thereby obtaining the data point clustering boundary information.

[0034] Based on the cluster boundary information of data points, a data distribution reference plane is constructed so that the data distribution reference plane accurately reflects the relative positional relationship of each data cluster in multidimensional space, resulting in a reference plane structure with geometric characteristics.

[0035] Based on a reference plane structure with geometric properties, a data processing base region is defined so that the data processing base region covers the distribution range of all data clusters and preserves the correlation between data, thus obtaining a complete base region definition;

[0036] By defining the complete base region and combining cluster boundary information and data distribution density characteristics, the base region is divided into multiple interconnected sub-regions. Each sub-region corresponds to specific data distribution characteristics and computational task types, forming the sub-region partitioning result.

[0037] Furthermore, based on the sub-region partitioning results, multidimensional data points are mapped to corresponding sub-regions, and task configuration parameters are obtained according to the aggregation characteristics of data points within each sub-region. These task configuration parameters are then used to adapt and adjust the computational task logic of each privacy computing node. The central coordination node then uses this adjusted task logic to obtain corresponding coordination instructions and task payloads, including:

[0038] Based on the sub-region division results, the multidimensional data points are mapped to the corresponding sub-regions according to their data feature attributes, thus obtaining the distribution status of data points in each sub-region.

[0039] Based on the distribution of data points in each sub-region, the data aggregation density, distribution uniformity, and boundary clarity characteristics of each sub-region are calculated to obtain quantitative indicators of data aggregation characteristics.

[0040] By quantifying indicators based on data aggregation characteristics, corresponding task configuration parameters are calculated and generated for each sub-region. The task configuration parameters include the proportion of computing resources allocated, task execution priority, and data processing granularity, forming a set of task configuration parameters.

[0041] By using the set of task configuration parameters, the original computing task logic of each privacy computing node is adapted and adjusted in a differentiated manner, so that the computing task logic of each node matches the data characteristics of the sub-region in which it is located, and thus the adapted and adjusted set of task logic is obtained.

[0042] Based on the adapted and adjusted task logic set, a coordinated instruction sequence and task payload are generated for each privacy computing node.

[0043] Furthermore, by coordinating the command flow and task payload flow for unified scheduling and distributing routes to the corresponding privacy computing nodes, each node is scheduled to execute computing tasks locally, achieving collaborative processing of computation localization and centralized management, including:

[0044] Based on the coordinated instruction sequence and task payload data packet, a unified scheduling framework is constructed to form centralized management data for instruction flow and payload flow;

[0045] Based on centralized management data, the coordinated instruction flow is time-sequenced according to task dependencies and execution priorities to obtain the instruction execution timing table;

[0046] Based on the instruction execution timing table, route planning is performed on the task payload flow to determine the target node path and distribution strategy of each payload data packet, and a route planning scheme is generated.

[0047] According to the routing plan, the coordination instructions and task payloads are distributed to the corresponding privacy computing nodes through the transparent forwarding mechanism of the central coordination node, forming a distribution execution state;

[0048] Based on the distributed execution status, the local computing execution process of each privacy computing node is monitored, computing status information fed back by the nodes is collected, and abnormal execution status is centrally coordinated and dynamically adjusted. This allows the central coordination node to be only responsible for instruction scheduling and status monitoring without participating in actual computing, ultimately realizing a collaborative architecture where computing tasks are executed on local nodes while the management process is centralized at the central node.

[0049] Secondly, a multi-terminal data privacy collaborative processing system based on a central network node includes:

[0050] The acquisition module is used to deploy multiple heterogeneous privacy computing nodes with different privacy computing protocols, network communication protocols, and upper-layer application designs.

[0051] The adjustment module is used to set up a central coordination node to address the need for interconnection and interoperability of heterogeneous privacy computing nodes. The central coordination node has protocol compatibility capabilities.

[0052] The connection module is used to enable each heterogeneous privacy computing node to establish a network connection with the central node through the protocol compatibility of the central node, and to establish an indirect interconnection channel between all privacy computing nodes through the central node.

[0053] The partitioning module is used to collect source data from each node based on the established interconnection channels and transform it into a multi-dimensional data point set. By performing cluster boundary analysis on the multi-dimensional data point set, the cluster boundary information of the data point set is obtained. Based on the cluster boundary information, a reference plane is constructed, the base region is defined, and interrelated sub-regions are divided according to the data distribution and cluster boundary information to form the sub-region partitioning result.

[0054] The adjustment module is used to map multidimensional data points to corresponding sub-regions based on the sub-region division results, obtain task configuration parameters based on the aggregation characteristics of data points in each sub-region, adapt and adjust the computing task logic of each privacy computing node based on the task configuration parameters, and obtain corresponding coordination instructions and task payloads by the central coordination node based on the adjusted task logic.

[0055] The processing module is used to coordinate the instruction stream and task payload stream for unified scheduling and distribute routes to the corresponding privacy computing nodes, so as to schedule each node to execute computing tasks locally, thereby achieving collaborative processing of computing localization and centralized management.

[0056] Thirdly, a computing device, comprising:

[0057] One or more processors;

[0058] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0059] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0060] The above-described solution of the present invention has at least the following beneficial effects:

[0061] By employing a combination of techniques—deploying multiple heterogeneous privacy computing nodes and setting up a central coordination node with full protocol compatibility; establishing indirect interconnection channels between heterogeneous nodes through the central node; dividing sub-regions based on data clustering boundary analysis and generating task configuration parameters adapted to data characteristics; and constructing a unified scheduling framework to achieve centralized scheduling of instructions and payloads, with the central node only responsible for routing and monitoring and not participating in actual computation—this approach effectively overcomes the technical problems of existing multi-terminal privacy computing nodes, such as complex networking and high connection costs due to protocol heterogeneity, mismatch between task allocation and data characteristics and waste of computing resources due to the lack of a unified scheduling mechanism, and the risk of data privacy leakage caused by the central node's participation in computation. This approach achieves the technical effects of breaking down barriers to interconnection between heterogeneous nodes, ensuring data privacy and security, improving computing resource utilization, realizing collaboration between localized computation and centralized management, and enhancing the efficiency and security of multi-terminal data privacy collaborative processing. Attached Figure Description

[0062] Figure 1 This is a flowchart illustrating a multi-terminal data privacy collaborative processing method based on a central network node, provided in an embodiment of the present invention.

[0063] Figure 2 This is a schematic diagram of a multi-terminal data privacy collaborative processing system based on a central networking node, provided by an embodiment of the present invention. Detailed Implementation

[0064] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0065] like Figure 1 As shown, embodiments of the present invention propose a multi-terminal data privacy collaborative processing method based on a central network node, the method comprising the following steps:

[0066] Step 1: Deploy multiple heterogeneous privacy computing nodes with different privacy computing protocols, network communication protocols, and upper-layer application designs;

[0067] Step 2: To address the need for interconnection and interoperability among heterogeneous privacy computing nodes, a central coordination node is set up, which has protocol compatibility capabilities.

[0068] Step 3: Through the protocol compatibility of the central node, each heterogeneous privacy computing node establishes a network connection with the central node, thereby establishing an indirect interconnection channel between all privacy computing nodes through the central node.

[0069] Step 4: Based on the established interconnection channels, the central node collects source data from each node and transforms it into a multi-dimensional data point set; by performing cluster boundary analysis on the multi-dimensional data point set, the cluster boundary information of the data point set is obtained; based on the cluster boundary information, a reference plane is constructed, the base region is defined, and interrelated sub-regions are divided according to the data distribution and cluster boundary information to form the sub-region division result.

[0070] Step 5: Based on the sub-region division results, the multi-dimensional data points are mapped to the corresponding sub-regions. The task configuration parameters are obtained according to the aggregation characteristics of the data points in each sub-region. The computation task logic of each privacy computing node is adapted and adjusted according to the task configuration parameters. The central coordination node obtains the corresponding coordination instructions and task payloads based on the adjusted task logic.

[0071] Step 6: By coordinating the instruction stream and task payload stream for unified scheduling, and distributing routes to the corresponding privacy computing nodes, each node is scheduled to execute computing tasks locally, thereby achieving collaborative processing of computing localization and centralized management.

[0072] In this embodiment of the invention, by deploying multiple heterogeneous privacy computing nodes and setting up a central coordination node with protocol compatibility, the central node enables each heterogeneous node to form a single network and establish an indirect interconnection channel. Based on this channel, the data collected by the central node is transformed into a multi-dimensional data point set, and then the partitioning results are obtained through cluster boundary analysis, reference plane construction, and sub-region division. Then, based on the data aggregation characteristics of the sub-regions, task configuration parameters are generated to adapt to the node computing task logic and generate coordination instructions and task payloads. Finally, the unified scheduling instruction stream and task payload stream are distributed to each node to achieve local computing and centralized management. This technical means effectively overcomes the technical problems of existing multi-terminal privacy computing nodes, such as difficulty in interconnection and high networking complexity caused by heterogeneous protocols and application architectures, unreasonable task allocation and waste of computing resources due to the lack of targeted data processing and unified scheduling mechanisms, and the risk of data privacy leakage caused by the participation of the central node in computing. Thus, it achieves the technical effects of breaking down the barriers to collaboration between heterogeneous nodes, ensuring data privacy and security, improving the utilization rate of computing resources, realizing efficient collaboration between localized computing and centralized management, and improving the overall efficiency and reliability of multi-terminal data privacy collaborative processing.

[0073] In a preferred embodiment of the present invention, step 1 above may include:

[0074] Step 1.1 involves performing heterogeneity identification and analysis on each privacy computing node to be connected, obtaining analysis results on differences in privacy computing protocol types, network communication protocol types, and upper-layer application design architectures. Specifically, this includes: establishing a node heterogeneity identification working group, clarifying a list of all privacy computing nodes to be connected, covering nodes deployed in different institutions, including nodes in different application scenarios such as medical institutions and research institutions; for each node to be connected, initiating a privacy computing protocol type identification process, using protocol feature scanning to sequentially detect the privacy computing protocol kernel currently running on the node, the protocol data encapsulation format, and the protocol encryption transmission process, recording whether the node uses RSA plus Hash, Chinese cryptographic standard plus Hash, ECDH, OPRF, or VOLE protocol, and marking the core differences between different node protocols, such as differences in encryption key length and data signature methods. The process involves identifying network communication protocol types. Through network packet capture and analysis, the communication interaction mode of each node is determined, identifying whether it uses a one-way network communication request-pull mode, a long-connection mode, or the gRPC protocol. Differences in transmission rate, connection persistence mechanism, and data packet segmentation strategy among different communication protocols are recorded. Next, the upper-layer application design architecture is identified. Through application architecture document parsing and application runtime monitoring, it is determined whether each node's upper-layer application uses a layered architecture, microservice architecture, or monolithic architecture. Differences in application interface naming rules, data interaction processes, and business logic module divisions are recorded. The results from these three aspects are then summarized to form a complete heterogeneous characteristic analysis report. This report clearly lists the differences in privacy computing protocol types, network communication protocol types, and upper-layer application design architecture for each node to be connected.

[0075] Step 1.2: Based on the heterogeneity characteristic identification and analysis results, configure privacy computing protocol compatibility capabilities to support the conversion of multiple privacy computing protocols, including RSA+Hash, Chinese cryptographic standard + Hash, ECDH, OPRF, and VOLE. Specifically, this includes: based on the generated heterogeneity characteristic analysis report, identifying the types of privacy computing protocols that need to be compatible, and determining that this configuration needs to cover five protocols: RSA+Hash, Chinese cryptographic standard + Hash, ECDH, OPRF, and VOLE; building a basic framework for the privacy computing protocol conversion engine, which has core functions such as protocol feature recognition, protocol rule mapping, protocol data conversion, and protocol compatibility verification; developing corresponding feature recognition rules for each protocol, and inputting the encryption features, data format features, and interaction flow features of different protocols in the protocol feature recognition stage, setting the protocol feature matching threshold to 80. When the matching degree between the node protocol features and the input features reaches 80 or higher, the protocol type is considered successfully identified. In the protocol rule mapping stage, bidirectional mapping rules are established between the five protocols, clarifying the correspondence between encryption algorithms, data fields, and signature verification mechanisms of different protocols. For example, mappings are established between the key conversion rules and data digest generation rules between Chinese cryptographic standard plus hash and RSA plus hash. In the protocol data conversion stage, data conversion logic is developed based on the mapping rules to achieve lossless conversion of data from one protocol format to another, ensuring that the converted data retains its original privacy protection characteristics and data integrity. The protocol compatibility verification process is initiated, and test nodes of different protocol types are selected to conduct cross-protocol data interaction tests to verify whether the converted data can be normally decrypted, verified, and processed by the target node. The protocol conversion rules are optimized based on the test results, and finally, the configuration of privacy computing protocol compatibility is completed.

[0076] Step 1.3: Based on the configuration results of privacy computing protocol compatibility, establish network communication protocol adaptation capabilities. This capability handles protocol conversions for single-transaction network communication (request-pull mode), long-connection mode, and gRPC protocol, resulting in the establishment of the network communication protocol adaptation capability. Specifically, this includes: determining that the network communication protocol adaptation capability needs to work in conjunction with the configured privacy computing protocol compatibility capability to ensure normal data transmission after protocol conversion; constructing a network communication protocol adaptation layer architecture with core functions of communication protocol identification, communication mode conversion, and communication quality monitoring; and in the communication protocol identification stage, recording the communication characteristics of single-transaction network communication (request-pull mode), long-connection mode, and gRPC protocol, such as request triggering characteristics for request-pull mode, connection heartbeat characteristics for long-connection mode, and HTTP / 2 transmission characteristics for gRPC protocol. The communication protocol identification response time is set to 1 second to ensure rapid identification of the communication protocol type accessed by the node. In the communication mode conversion phase, conversion logic between three communication protocols was developed. For the request-pull mode, a request-response timeout threshold of 30 seconds was set to ensure that the requester receives the response data within the specified time. For the long-connection mode, a heartbeat detection interval of 10 seconds was set; if no heartbeat signal is detected for three consecutive times, the connection is automatically disconnected and re-initiated. For the gRPC protocol, the message fragment size was set to 1024 kilobytes to ensure that large data volumes are fragmented and transmitted and completely reassembled at the receiving end. In the communication quality monitoring phase, a transmission rate monitoring threshold of 1 megabyte per second was set. When the data transmission rate is lower than this threshold, communication mode optimization and adjustment are automatically triggered, such as switching from request-pull mode to long-connection mode to improve transmission efficiency. Network communication protocol adaptation capability testing was conducted, simulating data transmission between nodes with different communication protocol types to verify whether the adaptation layer can achieve automatic protocol identification and conversion. At the same time, the stability and integrity of data transmission were monitored. Based on the test results, the adaptation parameters were adjusted to obtain the established network communication protocol adaptation capability.

[0077] Step 1.4: Based on the results of establishing network communication protocol adaptation capabilities, generate application interface standardization capabilities. This standardization unifies the application interface specifications of different heterogeneous privacy computing platforms, achieving standardized mapping of heterogeneous application design styles. The resulting application interface standardization capabilities include: Based on the network communication protocol adaptation capabilities, clarifying the core objective of application interface standardization capabilities: unifying the interface specifications of different heterogeneous privacy computing platforms, achieving seamless integration of interface interaction and communication protocol adaptation layers in financial scenarios, ensuring smooth data transmission and collaborative computing for customers, and complying with privacy protection regulations regarding customer information security; and developing unified application interface specifications covering the collaborative needs of financial business: interface naming uses a combination of business module and function name; request parameters are in key-value pair format, containing core information such as customer identifiers and risk indicator codes. The response data consistently includes status codes, message prompts, and business data. The call flow is: authentication first, then request submission, and finally response reception. Exception handling uniformly returns standard error codes and descriptions. A standardized application interface mapping system is built, featuring core functions such as interface specification parsing, difference comparison, and mapping conversion, ensuring processing latency does not exceed 100 milliseconds to adapt to high concurrency requirements. The parsing phase inputs unified specifications, parsing the naming, parameters, processes, and exception handling characteristics of existing interfaces at each node. The comparison phase clarifies differences in parameter names, types, call processes, and exception handling dimension by dimension. The conversion phase develops mapping rules based on these differences, achieving unified parameter names, automatic type conversion, process adaptation, and standardized error code mapping. An interface compatibility verification threshold of 90 is set, and the adaptation effect is judged by the average of parameter matching degree and process matching degree; if it is less than 90, customized mapping logic is developed. Financial scenario testing is conducted, covering typical processes such as customer credit data transmission, verifying an interface call success rate of no less than 99.5%, data transmission accuracy of no less than 99.9%, and a response time of no more than 300 milliseconds. Problems are recorded, mapping rules are optimized, and after repeated testing to meet the standards, the application interface standardization capability is generated.

[0078] Step 1.5: Based on the generated application interface standardization capabilities, integrate privacy computing protocol compatibility, network communication protocol adaptation, and application interface standardization capabilities to complete the unified deployment of multiple heterogeneous privacy computing nodes. Specifically, this includes: building a unified deployment platform for heterogeneous privacy computing nodes; integrating the configured privacy computing protocol compatibility, established network communication protocol adaptation, and generated application interface standardization capabilities to form a three-in-one node access and management capability; formulating a node deployment process, which includes four stages: node information entry, capability configuration, joint debugging and testing, and formal launch; in the node information entry stage, the heterogeneous characteristic analysis report of the nodes to be connected is entered into the deployment platform, establishing a unique identity for each node; in the capability configuration stage, based on the heterogeneous characteristics of the nodes, each node is configured with... Configure corresponding privacy computing protocol compatibility strategies, network communication protocol adaptation strategies, and application interface standardization strategies. For example, configure conversion strategies between national cryptographic hashing and other protocols for nodes using the national cryptographic hashing protocol, and configure adaptation strategies between requestpull and other communication modes for nodes using the requestpull communication mode. During the joint debugging and testing phase, organize nodes with the relevant strategies to conduct cross-node privacy computing collaborative tests. The test content includes the effectiveness of protocol conversion, communication adaptation stability, and interface call accuracy, while monitoring the computing performance and data privacy protection capabilities of the nodes. In the formal launch phase, the nodes that pass the joint debugging and testing are connected to the unified deployment platform and incorporated into the heterogeneous privacy computing node cluster for management, completing the unified deployment of multiple heterogeneous privacy computing nodes.

[0079] In this embodiment of the invention, by first identifying the heterogeneous characteristics of the privacy computing nodes to be accessed to clarify the differences in protocols, communication, and application architectures, and then specifically configuring compatibility capabilities to support the conversion of multiple privacy computing protocols and establishing adaptation capabilities for multiple types of network communication protocols, and subsequently generating application interface standardization capabilities and integrating the three types of capabilities to complete the unified deployment of nodes, the technical means effectively overcome the technical problems of existing heterogeneous privacy computing nodes having high access adaptation difficulty and high networking collaboration threshold due to inconsistent protocol types, communication modes, and application interfaces, as well as the inability of single capability configuration to cover multi-dimensional heterogeneous differences. In this way, it achieves full compatibility and standardization of heterogeneous nodes at the protocol, communication, and application interface levels, and reduces the cost of node access and networking collaboration.

[0080] In a preferred embodiment of the present invention, step 2 above may include:

[0081] Step 2.1, based on the networking requirements for interconnection of heterogeneous privacy computing nodes, construct the basic architecture of the central coordination node. This includes: conducting a survey on the networking requirements of heterogeneous privacy computing nodes, collecting data on the number of nodes, geographical distribution, computing resource configuration, data interaction frequency, and protocol compatibility requirements of different institutions, and clarifying the concurrent connection count, data transmission rate, and protocol conversion capability indicators that the central coordination node needs to support; designing the hardware architecture of the central coordination node, adopting a multi-node cluster deployment mode, configuring high-performance servers as core processing nodes, with each core node equipped with a 16-core CPU, 128GB of memory, and 10TB of storage, while deploying load balancing equipment to distribute requests, and configuring redundant backup equipment to ensure architecture stability; and building the software infrastructure, selecting stable operating systems. The system serves as the operating environment, deploying a distributed task scheduling system, data transmission middleware, and protocol processing engine. It is divided into four functional areas: a protocol compatibility area, a task scheduling area, a data storage area, and a status monitoring area, clearly defining the resource allocation ratio and data flow path for each area. An architecture expansion mechanism is designed to support the dynamic addition and removal of core processing nodes. When the number of concurrent connections exceeds 1000, node expansion is automatically triggered; when the number of connections falls below 500, the number of nodes is automatically reduced, ensuring the architecture can adapt to different network scale requirements. Architecture security is strengthened by deploying firewalls, intrusion detection systems, and data encryption transmission components to encrypt communication data between nodes throughout the entire process. Access control policies are set to allow only authorized heterogeneous privacy computing nodes to access the system, completing the construction of the central coordination node infrastructure.

[0082] Step 2.2: Based on the infrastructure of the central coordination node, configure the network protocol compatibility layer to identify and convert communication data in one-way network communication request-pull mode, long connection mode, and gRPC protocol, obtaining the configuration result of the network protocol compatibility layer. Specifically, this includes: planning the deployment location of the network protocol compatibility layer based on the infrastructure of the central coordination node, integrating it between the data transmission middleware and the external node access port, ensuring that all communication data from heterogeneous nodes must be processed through this layer; developing a communication protocol identification function, collecting typical data packet samples of one-way network communication request-pull mode, long connection mode, and gRPC protocol using a network packet capture tool, extracting feature information such as packet header format, field identifiers, and data delimiters for each protocol, and establishing a protocol feature library; designing protocol identification logic, using a feature matching algorithm to perform real-time analysis of the input communication data, and matching data features with protocol features. The system compares information in the database, sets a protocol identification accuracy threshold of 95%, and considers successful identification when the matching degree reaches 95% or higher, outputting the corresponding protocol type. It develops a protocol conversion function, defining conversion rules for the differences between the three protocols. This converts request-pull mode request-response data into a standard data frame format, splits long-connection mode streaming data into fixed-length segments and adds frame identifiers, and parses and repackages gRPC protocol HTTP2 transmission data into a common format for the central node. It sets protocol conversion performance metrics, specifying a conversion latency threshold of 50 milliseconds and a data conversion accuracy threshold of 99.9%, ensuring that the conversion process does not affect the real-time performance and integrity of data transmission. Finally, it conducts protocol compatibility layer testing, simulating heterogeneous nodes sending communication data with different protocol types to verify the accuracy of protocol identification and the effectiveness of the conversion function. The conversion latency and data integrity during the test are recorded, yielding the configuration results of the network protocol compatibility layer.

[0083] Step 2.3: Based on the configuration results of the network protocol compatibility layer, construct a privacy computing protocol conversion capability. This capability supports real-time conversion and adaptation of multiple privacy computing protocols, including RSA+Hash, Chinese national cryptography+Hash, ECDH, OPRF, and VOLE. The specific steps include: 1) Based on the network protocol compatibility layer configuration results, clarify that the privacy computing protocol conversion capability must coordinate with the network protocol conversion process, receiving the data stream after network protocol conversion to ensure seamless data transmission and protocol conversion; 2) Review the technical details of five privacy computing protocols: RSA+Hash, Chinese national cryptography+Hash, ECDH, OPRF, and VOLE, including encryption algorithm principles, key generation rules, data encapsulation formats, signature verification processes, and interaction command sets, forming protocol technical documentation; 3) Establish a protocol mapping rule base, formulating bidirectional mapping rules for conversion scenarios between each protocol, clarifying the correspondence of encryption algorithms, key length adaptation standards, data field mapping methods, and signature verification. The mechanism's conversion logic, for example, converts a 256-bit key of the Chinese cryptographic protocol plus hash protocol to a 2048-bit key of the RSA plus hash protocol, ensuring that the converted key meets the security requirements of the target protocol. A protocol conversion processing flow is developed, firstly identifying the protocol type of the input data, then calling the corresponding mapping rules based on the identification result, decrypting the encrypted part of the data, re-encrypting and encapsulating it according to the target protocol format, generating the converted protocol data, while retaining the original business information and privacy protection characteristics of the data. A protocol conversion quality threshold is set, specifying a conversion power threshold of 99, ensuring that the privacy protection strength of the converted data is not lower than that of the original data, and ensuring that the conversion process does not reduce data security. Protocol conversion capability testing is conducted by selecting heterogeneous nodes using different privacy computing protocols for cross-protocol data interaction, verifying whether the converted data can be normally decrypted, processed, and verified by the target node, monitoring the conversion process's time consumption and resource usage, and optimizing the mapping rules and conversion process based on the test results to obtain the construction result of the privacy computing protocol conversion capability.

[0084] Step 2.4: Based on the construction results of the privacy computing protocol conversion capability, establish an application layer interface standardization mechanism. This mechanism will uniformly map the application interface specifications of different heterogeneous privacy computing platforms into a standard interface format recognizable by the central coordination node. Specifically, this includes: based on the construction results of the privacy computing protocol conversion capability, clarifying the core positioning of the application layer interface standardization mechanism; accurately receiving financial data streams converted by the privacy computing protocol to achieve seamless integration with the protocol conversion process; and simultaneously uniformly converting the application interface specifications of various heterogeneous privacy computing platforms into a format recognizable by the central coordination node to ensure smooth data interaction between multiple nodes and the central node in financial scenarios. Throughout the process, the system adheres to privacy regulations, ensuring no disclosure of customer financial information. A standardized interface specification for the central coordination node has been developed to meet core needs such as joint financial risk modeling. Interface naming uses a combination of functional modules and operation names to ensure intuitiveness and business relevance. Request parameters support multiple types, including core information such as customer identifiers and risk indicator codes, with clearly defined mandatory attributes and value ranges. Response data consistently includes three parts: status identifier, processing result, and business data. The call flow involves authentication first, then parameter submission, and finally receiving the response. Authentication is achieved through node identifiers and keys. Exception handling uniformly returns standard error codes and descriptions for quick problem localization. Comprehensive collection of application interface documents from various heterogeneous platforms has been conducted, with detailed analysis of interface naming conventions. The document outlines the differences between heterogeneous platform interfaces, including their syntax, parameter definitions, data interaction formats, and calling logic. It extracts core differences such as parameter names, data types, formats, and processes to create a list of interface differences. An interface mapping rule system is established to achieve a one-to-one accurate mapping between heterogeneous interfaces and the central standard interface, ensuring consistent business meaning and parameter correspondence at the parameter name level. At the data format level, various encapsulated formats are converted to a structured format supported by the central node to guarantee data accuracy. At the calling process level, adaptation logic such as temporary caching is used to ensure compatibility with non-standard process nodes. An interface conversion processing flow is developed, and a conversion module is built: parsing heterogeneous node requests to extract key information, matching mapping rules, converting to a standard format, and submitting to the central node; and converting the response received from the central node according to the reverse mapping rules. To ensure the heterogeneous nodes can recognize the format, core financial data and privacy attributes remain unchanged throughout the process. Standardized verification thresholds were set for the interface: conversion power 98, interface call response time 200 milliseconds, supplemented with data conversion accuracy 99.9% and interface compatibility 95%, guaranteeing the mechanism's universality and data accuracy. Interface standardization testing was conducted, simulating typical business processes in financial scenarios, selecting multiple heterogeneous financial privacy computing nodes to interact with the central node. Core indicators such as conversion power and response time were monitored to verify the stability of concurrent access from 100 nodes. Mapping rules and conversion processes were optimized to address issues such as parameter mapping errors and format conversion anomalies, with repeated testing until all indicators met the standards, ultimately resulting in the establishment of the application-layer interface standardization mechanism.

[0085] Step 2.5, leveraging the established application layer interface standardization mechanism, integrates the network protocol compatibility layer, privacy computing protocol conversion capabilities, and the application layer interface standardization mechanism to form a complete protocol compatibility capability for the central coordination node. Specifically, this includes: designing a three-layer capability integration architecture, connecting the network protocol compatibility layer, privacy computing protocol conversion capabilities, and the application layer interface standardization mechanism in the order of data flow, clarifying the input / output data formats, call triggering conditions, and exception handling methods for each layer, forming a unified capability call process; developing a capability collaborative scheduling process, which receives access requests from heterogeneous nodes, automatically calls the corresponding processing capabilities based on the request type, first performs communication protocol conversion through the network protocol compatibility layer, then performs protocol adaptation through the privacy computing protocol conversion capability, and finally completes interface format unification through the application layer interface standardization mechanism, ensuring that data enters the central coordination node after sequential processing through the three layers; establishing a data flow monitoring mechanism to monitor the data status in real time during the three-layer capability processing, recording the data's protocol type, conversion result, interface format, and processing time, and setting... A data processing anomaly threshold is established. When the processing time for a single piece of data exceeds 500 milliseconds or the conversion fails, a retry mechanism is automatically triggered, with a maximum of 3 retries. If the conversion still fails, the error message is recorded and relevant maintenance personnel are notified. Capability extension and adaptation rules are formulated to support the rapid access of new network communication protocols, privacy computing protocols, and application interface specifications. When compatibility with new protocols or interfaces is required, only the corresponding feature library, mapping rules, and conversion logic need to be supplemented, without modifying the overall integration architecture. Complete protocol compatibility testing is conducted by selecting multiple heterogeneous privacy computing nodes using different protocols and interface specifications to simulate actual networking scenarios for access testing. This verifies whether the central coordination node can correctly identify protocol types, complete conversion processing, and interface adaptation. The processing performance and stability when multiple nodes access concurrently are monitored, and the success rate and latency of cross-node data interaction are tested. Based on the test results, the collaborative scheduling logic and processing flow at each level are optimized, and relevant threshold parameters are adjusted to ensure that the central coordination node efficiently and stably handles the access and data interaction needs of various heterogeneous nodes, ultimately forming the complete protocol compatibility capability of the central coordination node.

[0086] In this embodiment of the invention, by adopting a central coordination node infrastructure built based on the interconnection requirements of heterogeneous nodes, and then sequentially configuring a network protocol compatibility layer that can identify and convert multiple network communication protocols, building the ability to support real-time conversion of multiple privacy computing protocols, and establishing a standardized mechanism that can uniformly map heterogeneous application interfaces, the technical means of integrating three core capabilities to form the complete protocol compatibility capability of the central coordination node effectively overcomes the technical problems of existing central nodes lacking a layered protocol compatibility architecture, being unable to simultaneously adapt to multiple types of network communication and privacy computing protocols, and being difficult to be compatible with heterogeneous application interfaces, resulting in the inability to efficiently connect to heterogeneous privacy computing nodes and a weak foundation for network collaboration. Thus, the technical effect of enabling the central coordination node to have full-dimensional protocol compatibility and interface adaptation capabilities, achieving accurate and efficient connection with various heterogeneous nodes, and reducing the overall difficulty of network collaboration is achieved.

[0087] In a preferred embodiment of the present invention, step 3 above may include:

[0088] Step 3.1: Utilizing the complete protocol compatibility capabilities of the central coordination node, initialize the network service of the central node to provide a unified access endpoint for heterogeneous privacy computing nodes, thus obtaining the network service initialization status. This specifically includes: firstly, performing a pre-verification of the complete protocol compatibility capabilities of the central coordination node to confirm that the network protocol compatibility layer, privacy computing protocol conversion capabilities, and application layer interface standardization mechanisms are all in normal operating condition, and that various conversion thresholds, recognition accuracy, and other indicators meet preset requirements; secondly, initiating the network service initialization process by calling the service management program within the central node to configure the core parameters of the network service, including the service listening port, maximum concurrent access count, and connection timeout time. The maximum concurrent access count is set to 2000, and the connection timeout time is set to 60 seconds to avoid resource consumption due to prolonged connection waiting; based on the complete protocol compatibility capabilities... Configure the network attributes of the unified access endpoint, assign a fixed public IP address and dedicated port, and establish a connection between the access endpoint and the protocol compatibility processing flow of the central node to ensure that all network requests accessing this endpoint can directly enter the protocol identification and conversion stage; deploy an access request queuing mechanism, which automatically starts a queuing queue when the number of concurrent accesses approaches the threshold of 2000, with a queue length of 500. Access requests exceeding the queue length will return a queuing prompt to avoid service overload; perform availability testing on the initialized network service, simulating heterogeneous privacy computing nodes of different protocol types sending access probe requests to verify the response time and request reception success rate of the unified access endpoint. The response time should not exceed 100 milliseconds, and the reception success rate should not be less than 99.9%. After the test is passed, record the network service initialization status as normal and complete the initialization process.

[0089] Step 3.2: Based on the initialization state of the network service, receive network connection requests from each heterogeneous privacy computing node. Perform network protocol identification and conversion processing on each network connection request to obtain a standardized network request. Specifically, this includes: starting the access request receiving process based on the normal initialization state of the network service, monitoring the network port of the unified access endpoint in real time, capturing network connection request data sent by all heterogeneous privacy computing nodes, and recording the request sending time and the preliminary identification information of the sending node; performing legality verification on each received network connection request, including the integrity of the request data and the pre-authorization status of the sending node, allowing only requests from nodes that have completed registration and filing at the central node to pass the verification, and directly rejecting requests from unauthorized nodes and recording them in the log; starting the network protocol identification process, comparing the characteristics of the request data with a preset protocol feature library, which contains individual network features. The core characteristics of communication request-pull mode, long connection mode, and gRPC protocol packet header format, field identifiers, etc., are identified. A protocol identification accuracy threshold of 95% is set. Successful identification is determined when the matching degree reaches 95% or higher, confirming the network protocol type corresponding to the request. For the identified protocol type, the conversion function of the network protocol compatibility layer is invoked. The original connection request is converted according to the unified network request format within the central node. The conversion includes standardization of data frame structure, unification of field names, and unification of data encoding format, generating a standardized network request. The standardized network request undergoes format verification to check if it meets the preset requirements for the number of fields, field types, and data length. A format verification pass rate threshold of 99.5% is set. Standardized network requests that pass verification proceed to subsequent processing; requests that fail return a conversion failure message, requiring the node to resend.

[0090] Step 3.3: Based on the converted standardized networking request, establish a single-connection network between each heterogeneous privacy computing node and the central node, obtaining a node-to-central connection mapping table. Specifically, this includes: for a successfully verified standardized networking request, the central coordinating node sends a connection establishment response to the corresponding heterogeneous privacy computing node. The response data includes the central node's authentication information and temporary session identifier. The central node receives connection confirmation information from the heterogeneous privacy computing node, verifies the temporary session identifier, and, upon successful verification, initiates the single-connection network establishment process. A point-to-point connection method is used, allocating independent connection resources to each node to avoid resource contention between nodes. During the connection establishment process, the core information of each node is recorded, including the node's unique identifier, the privacy technology used, etc. The system identifies the private computing protocol type, the converted network protocol type, the connection establishment time, and the port resources used by the connection, while assigning a unique connection identifier to each connection. A node-to-central connection mapping table is constructed, containing fields such as unique node identifier, connection identifier, private computing protocol type, network protocol type, connection status, connection establishment time, and timeout. Connection status is categorized into three types: normal, disconnected, and pending reconnection. A dynamic update mechanism for the mapping table is established to monitor the operational status of each connection in real time. When a connection is disconnected, the corresponding record's connection status is automatically updated to disconnected, and a reconnection prompt is triggered. When a connection times out and remains inactive, the timeout is set to 300 seconds, and the status is updated to pending reconnection, ensuring that the mapping table accurately reflects the connection status between nodes and the central node.

[0091] Step 3.4: Based on the node-central connection mapping table, construct the global routing table for the central node. This allows the global routing table to forward data streams from any source node to the corresponding target node via the central node, based on the target node identifier. This results in the central node's routing and forwarding capabilities. Specifically, this includes: establishing a standardized application layer interface mechanism based on privacy computing protocol conversion capabilities; accurately receiving financial data streams after protocol conversion; unifying the application interface specifications of various heterogeneous privacy computing platforms into a format recognizable by the central coordination node; ensuring smooth data interaction between multiple nodes and the central node in financial scenarios; fully complying with privacy protection regulations; and strictly preventing the leakage of customer financial information. Furthermore, it involves developing standard interface specifications for the central coordination node to fully adapt to financial business needs: interface naming uses a combination of functional modules and operation names to ensure intuitiveness and business relevance; request parameters support integer, string, and array types, with core parameters including key information such as customer identifiers and risk indicator codes. Clearly define the attributes and value ranges of each parameter; the response data should consistently include three parts: status identifier, processing result, and business data; the call process involves first completing node authentication before submitting the request, with authentication achieved through the node's unique identifier and key; exception handling uniformly returns standard error codes and descriptions for quick problem location; collect application interface documents from various heterogeneous platforms, meticulously analyze interface naming, parameter definitions, data interaction formats, and call logic, extract core differences such as parameter names, data formats, and call processes, and form a list of heterogeneous platform interface differences; establish an interface mapping rule system to achieve a one-to-one accurate mapping between heterogeneous interfaces and the central standard interface, clarifying parameter correspondence, format conversion methods, and process adaptation logic to ensure financial data format compatibility; develop an interface conversion process to parse heterogeneous node requests and convert them into a standard format according to rules before submitting them to the central node, and after receiving the central response, convert it back to a format recognizable by the heterogeneous node according to reverse rules, without altering core financial data and privacy attributes throughout the process. The verification thresholds were set, with an interface conversion power of 98, a call response time of 200 milliseconds, supplementary data conversion accuracy of 99.9%, and interface compatibility of 95, ensuring the efficiency and universality of the mechanism. Interface standardization testing was carried out, simulating financial business scenarios such as customer credit data transmission, allowing various heterogeneous platforms to interact with the central node through their own interfaces, monitoring core indicators, verifying the stability of multi-node concurrency, and after repeated testing to meet the standards, the results of the application layer interface standardization mechanism were obtained.

[0092] Step 3.5: Activate the transparent data flow forwarding mechanism of the central node through its routing and forwarding capabilities. This enables the central node to only perform data routing and distribution without participating in actual computation, establishing an indirect interconnection channel between all privacy computing nodes. Specifically, this includes: initiating the activation process of the transparent data flow forwarding mechanism based on the central node's routing and forwarding capabilities. The core objective is to achieve efficient forwarding of financial data based on routing and forwarding, while ensuring that the central node only undertakes the function of routing and distribution without participating in actual computation, thus meeting the privacy protection requirements of financial scenarios. Configure core parameters for the forwarding mechanism: set the data forwarding buffer size to 10GB to accommodate large-volume financial data transmission; set the forwarding latency threshold to 50 milliseconds, automatically triggering buffer expansion when the forwarding latency exceeds this threshold; set the data discard threshold to 100 milliseconds, allowing data to be discarded only if it exceeds this threshold and fails to complete forwarding, and recording the discard time, data identifier, and other information in detail; clearly define transparent forwarding core rules: the central node only routes and distributes financial data streams and does not participate in any data decryption, calculation, or modification operations; maintain the original encrypted state and data structure during forwarding, adding only a temporary forwarding identifier to the data header, which contains source node, target node, and forwarding path information for the central node to identify the forwarding direction; automatically remove the identifier after forwarding to avoid affecting the original attributes of the data; allocate independent buffers for forwarded data between privacy computing nodes of different financial institutions to prevent interference between different data streams. Simultaneously, a real-time data forwarding monitoring process was initiated to continuously record data forwarding status information, such as whether the data was pending forwarding, forwarding failed, forwarding time, and transmission rate, ensuring traceability of the entire forwarding process and facilitating subsequent troubleshooting. Interconnectivity tests between nodes were conducted, selecting heterogeneous financial privacy computing nodes with various protocol types. Nodes were instructed to send financial test data through the central node to verify whether the data could be forwarded normally and whether the target nodes could accurately receive and parse it. The central node's operational status was simultaneously checked to confirm it was not involved in data computation. A large-volume financial risk data transmission scenario was simulated to verify whether the buffer capacity met the requirements and whether the forwarding latency was controlled within the 50-millisecond threshold, ensuring the integrity and real-time nature of data transmission. After all test items met the standards, it was confirmed that all financial privacy computing nodes could achieve indirect data interaction through the central node's transparent forwarding mechanism, completing the establishment of the indirect interconnectivity channel.

[0093] In this embodiment of the invention, by employing the full protocol compatibility capability of the central coordinating node to initialize the networking service and provide a unified access endpoint, and by receiving networking connection requests from heterogeneous privacy computing nodes, performing protocol identification and conversion, establishing a single networking connection for each node and generating a node-central connection mapping table, constructing a global routing table based on this table, activating the central node's transparent data flow forwarding mechanism, and thus establishing an indirect interconnection channel for all privacy computing nodes, this technical approach effectively overcomes the existing technical problems of high protocol adaptation costs and high networking complexity for heterogeneous privacy computing nodes, the risk of privacy leakage due to direct data interaction between nodes, and the lack of a global routing management mechanism leading to low data flow forwarding efficiency and poor directionality. This significantly reduces the difficulty and cost of heterogeneous node networking adaptation, enabling full network interconnection with a single connection between nodes and the central node. The central node is only responsible for data routing and distribution and does not participate in actual computation, ensuring data privacy and security. Simultaneously, the global routing table enables precise targeted forwarding of data flows, improving the efficiency of data interaction between nodes.

[0094] In a preferred embodiment of the present invention, step 4 above may include:

[0095] Step 4.1 involves collecting source data feature information from various privacy computing nodes through interconnection channels and transforming it into a standardized set of multi-dimensional data points. Specifically, this includes: initiating the data collection process based on the established indirect interconnection channels, defining the collection scope as source data feature information related to joint financial risk modeling within the privacy computing nodes of various financial institutions, covering core financial data dimensions such as customer credit characteristics, transaction behavior characteristics, credit history characteristics, debt characteristics, and risk indicator characteristics; employing encrypted transmission links throughout the collection process, strictly adhering to relevant privacy protection regulations to ensure that customer financial information is not leaked and to guarantee the security and privacy of data transmission; and preprocessing the collected source data feature information, firstly performing data integrity verification, removing invalid data lacking key financial features, setting a data integrity threshold of 95, and retaining data when the feature integrity of a single data point reaches 95 or higher. Simultaneously, the system identifies and corrects abnormal data. For a small number of missing non-critical financial feature data, mean imputation is used to supplement them, comprehensively improving data quality and laying the foundation for subsequent standardization conversion. Unified data standardization rules are established, clarifying the coding methods, dimensional uniformity standards, and data format requirements for various financial features. The value range of all feature data is normalized to the 0-1 interval, ensuring the comparability and uniformity of various financial features. According to the standardization rules, the preprocessed source data features are transformed, mapping multiple features of each financial data point to a data point in a multi-dimensional space. Each dimension corresponds to a standardized financial feature indicator, forming an initial multi-dimensional data point set. The initial data point set is then standardized and verified, calculating the standardization error for each data point. A standardization error threshold of 0.01 is set; when the error is less than 0.01, the standardization is considered successful. For unsuccessful data points, the conversion process is re-checked and a second conversion is performed until all data points meet the standardization requirements, ultimately resulting in an accurate and standardized multi-dimensional data point set.

[0096] Step 4.2: Based on the standardized representation of the multidimensional data point set, execute the point set clustering boundary extraction algorithm to perform density distribution analysis and boundary identification on the multidimensional data points, obtaining the data point clustering boundary information. Specifically, this includes: preprocessing the standardized multidimensional data point set to remove isolated noise points; setting a noise point judgment threshold of less than 0.05 for local data point density, identifying and removing data points with local densities below this threshold as noise points to avoid interfering with the clustering results; initializing the core parameters of the point set clustering boundary extraction algorithm, setting the neighborhood radius to 0.1 to determine the neighborhood range of each data point; setting the core point density threshold to 10, identifying a data point as a core data point when its neighborhood contains 10 or more data points; and executing the density distribution analysis process, traversing all data points and calculating... Based on the local density and distance of each data point to other core data points, different data clusters are formed. Core data points within the same data cluster are interconnected, while core data points in different data clusters are independent. For each data cluster, non-core data points located at the cluster edges are selected. The neighborhood of these data points contains data points from both the cluster and non-cluster data points or blank areas. By connecting the edge data points, closed boundaries are formed, obtaining preliminary cluster boundary information. The accuracy of the cluster boundaries is verified by calculating the cluster affiliation consistency of data points within the boundary. A boundary identification accuracy threshold of 98 is set, and the boundary is considered valid when the consistency reaches 98% or higher. For invalid boundaries, the neighborhood radius and core point density threshold are readjusted, and the clustering and boundary extraction process is repeated until accurate cluster boundary information of the data point set is obtained.

[0097] Step 4.3: Based on the clustering boundary information of the data points, construct a data distribution reference plane to accurately reflect the relative positional relationships of each data cluster in multidimensional space, resulting in a reference plane structure with geometric characteristics. Specifically, this includes: organizing the clustering boundary information of the data points; extracting the coordinates of the core cluster center of each data cluster, the distribution range of data points within the cluster, and the relative distances between clusters; using this information as the basic data for constructing the reference plane; ensuring that the dimension of the reference plane is consistent with the dimension of the multidimensional data points to ensure a complete mapping of the distribution state of the data points in multidimensional space; selecting a plane fitting method based on the core cluster center to construct the basic framework of the reference plane; and using the core cluster center of each data cluster as the key control point. The initial structure of the reference plane is obtained by substituting the data into the plane fitting model. By adjusting the plane parameters, the reference plane is made to fit the distribution trend of each data cluster to the maximum extent, accurately reflecting the relative positional relationship of different data clusters in multidimensional space. The fitting effect of the reference plane is verified by calculating the fitting error from the core point of each data cluster to the reference plane. The fitting error threshold is set to 0.02. When the fitting error of all core points is less than 0.02, the reference plane is deemed to be qualified. If there are core points with excessive errors, the fitting parameters are re-optimized and the fitting process is executed again. The qualified reference plane is structurally solidified by recording the dimensional parameters, fitting equation coefficients and key control point coordinates of the reference plane, forming a reference plane structure with geometric characteristics.

[0098] Step 4.4: Based on the reference plane structure with geometric characteristics, define the data processing base region. This base region should cover the distribution range of all data clusters while preserving the correlation between data points, resulting in a complete base region definition. Specifically, this includes: determining the definition range of the base region based on the reference plane structure with geometric characteristics; extending the base region in all directions of multidimensional space using the reference plane as a reference; ensuring the extension range completely covers the distribution range of all data clusters; setting an extension distance threshold of 0.2 to ensure the extended region encompasses all data points; analyzing the correlation between data clusters; identifying the correlation strength between data clusters by calculating the correlation coefficient between the core points of different data clusters; setting a correlation coefficient threshold of 0.6; and classifying clusters with a correlation coefficient greater than 0.6 as strongly correlated clusters. This process is then used to define the base region. The process involves preserving spatial connectivity between clusters; drawing the initial boundary of the base region based on the cluster boundary of the outermost data cluster, combined with the extension distance threshold and inter-cluster correlation requirements, to outline the approximate contour of the base region; ensuring that the contour covers all data clusters without including too many blank areas; verifying the coverage integrity and correlation preservation of the base region by checking whether all data clusters are completely located within the base region, setting the coverage integrity threshold to 100 to ensure that no data clusters are missed; simultaneously verifying whether strongly correlated clusters maintain spatial connectivity without obvious barriers to ensure that the correlation between data is not lost; optimizing and adjusting the boundary of the initial base region to correct irregular protrusions or depressions, making the region boundary smooth and continuous; finally, determining the specific range and boundary parameters of the base region to form a complete base region definition.

[0099] Step 4.5: Through the complete definition of the base region, combined with cluster boundary information and data distribution density characteristics, the base region is divided into multiple interconnected sub-regions. Each sub-region corresponds to specific data distribution characteristics and computational task types, forming the sub-region division results. Specifically, this includes: first, organizing the complete base region definition, data point set cluster boundary information, and data distribution density characteristic data; clarifying the core division criteria, using cluster boundaries as a natural separation reference and financial data distribution density characteristics as the core division principle, ensuring that each sub-region corresponds to specific financial risk data distribution characteristics, such as credit risk customer data characteristics, fraud risk transaction data characteristics, etc., while matching the corresponding computational task types, such as specific risk feature extraction tasks, risk level classification, etc., to meet the business needs of joint financial risk modeling; setting key thresholds for sub-region division, with the minimum data volume threshold for each sub-region set to 50, to ensure that each sub-region has sufficient data points to support subsequent financial risk-related computational tasks and avoid deviations in computational results due to insufficient data volume. Density partitioning thresholds are set based on data distribution density. Regions with a density higher than 0.8 are classified as high-density sub-regions, corresponding to scenarios such as high-frequency transaction risk data; regions with a density between 0.3 and 0.8 are classified as medium-density sub-regions; and regions with a density lower than 0.3 are classified as low-density sub-regions. Different density sub-regions are matched with different computational processing granularities: high-density regions use fine-grained processing to improve the accuracy of risk identification, while low-density regions use coarse-grained processing to ensure computational efficiency. Sub-region partitioning is performed by initially separating the base region along the cluster boundaries, while dynamically adjusting the separation boundaries based on data distribution density to ensure that data points within each sub-region have similar financial risk distribution characteristics. For strongly correlated adjacent data clusters, such as clusters related to credit risk and debt characteristics, the connectivity interfaces between sub-regions are preserved during partitioning to ensure that the correlation characteristics between financial data are not severed. The rationality of the partitioned sub-regions is verified by checking whether each sub-region meets the minimum data volume requirement, whether it corresponds to a clear financial risk data distribution characteristic and computational task type, and whether there is any overlap or omission between sub-regions. The sub-region partitioning pass rate threshold is set to 99%. When the pass rate of all sub-regions reaches 99% or above, the partitioning is deemed valid. For unqualified sub-regions, adjustments and optimizations are implemented, such as merging sub-regions with less than 50 data points, splitting overlapping sub-regions, and supplementing missing blank areas. After multiple rounds of optimization, the final partitioning result containing multiple interrelated sub-regions is formed.

[0100] In this embodiment of the invention, by collecting source data feature information from each node through interconnection channels and converting it into a standardized multidimensional data point set, executing a point set clustering boundary extraction algorithm to analyze data density distribution and identify boundaries, constructing a data distribution reference plane based on clustering boundary information and defining a base region covering all data clusters, and combining clustering boundaries and data distribution density features to divide corresponding sub-regions associated with specific data distribution features and computational task types, the technical means effectively overcome the existing problems of lack of standardized representation of distributed source data, difficulty in intuitively presenting the spatial distribution relationship of multidimensional data, and lack of data feature support for sub-region division leading to non-targeted subsequent task allocation. Thus, the standardized and unified processing of distributed data is achieved, accurately presenting the relative positional relationship of each data cluster in multidimensional space, and the divided sub-regions are highly matched with data features and computational task types.

[0101] In a preferred embodiment of the present invention, step 5 above may include:

[0102] Step 5.1: Based on the sub-region division results, map the multidimensional data points to the corresponding sub-regions according to their data feature attributes to obtain the data point distribution status within each sub-region. Specifically, this includes: first, reviewing the sub-region division results and clarifying the core financial data feature attributes corresponding to each sub-region. For example, some sub-regions correspond to credit risk customer characteristics, while others correspond to fraud risk transaction characteristics. Each sub-region clearly marks the feature attribute range, including customer credit score range, transaction amount range, credit history duration range, risk indicator coding range, etc., and compiles a list of sub-region feature attributes to provide a clear basis for data point mapping; second, perform secondary preprocessing on the standardized multidimensional financial data point set, focusing on removing a small amount of noise data remaining after the initial clustering. A noise data judgment threshold is set when the deviation between the data point feature and the core feature of its cluster is greater than 0.05. By calculating the deviation value between each data point and the corresponding cluster core feature one by one, data points with deviations exceeding the threshold are completely removed to avoid... To ensure the accuracy of subsequent mapping and guarantee the quality of input data, a data point feature matching process is initiated. Each multi-dimensional data point is sequentially traversed, extracting the values ​​of various financial characteristic attributes contained within the data point, including customer credit score, transaction frequency, debt amount, and credit history duration. The extracted feature values ​​are compared dimension-by-dimensionally with the feature range of each sub-region in the sub-region feature attribute list. The feature matching degree is calculated to determine the most suitable sub-region for each data point, ensuring that the matching results align with the distribution patterns of financial risk data. Based on the matching results, each multi-dimensional data point is accurately mapped to its corresponding sub-region. Key information such as the number of data points in each sub-region, the value distribution of each financial characteristic dimension, and the spatial coordinates of the data points within the sub-region are recorded simultaneously, integrating these elements to form the initial data point distribution state for each sub-region. The accuracy of the data point mapping is verified by calculating the average feature matching degree of all data points in each sub-region, setting the mapping accuracy threshold to 98%. When the average feature matching degree of data points in a certain sub-region reaches 98 or above, the sub-region is deemed to be qualified for mapping. For sub-regions where the average matching degree does not meet the standard, the feature comparison process of all data points in the region is re-checked, the matching deviation is corrected, and the mapping operation is re-executed until all sub-regions meet the mapping accuracy requirements, and finally the accurate distribution status of data points in each sub-region is obtained.

[0103] Step 5.2: Based on the data point distribution within each sub-region, calculate the data aggregation density, distribution uniformity, and boundary clarity characteristics of each sub-region to obtain quantitative indicators of data aggregation characteristics. Specifically, this includes: extracting the necessary basic data for calculation based on the data point distribution within each sub-region, including the boundary parameters of each sub-region, the coordinate information of the data points within the region, and the number of data points; calculating the data aggregation density of each sub-region by dividing the total number of data points within the sub-region by the volume of the sub-region in multidimensional space; setting the effective range of data aggregation density to 0 to 1, and re-verifying the calculation process of the number of data points and the region volume for outliers outside this range to ensure accurate results; and calculating the distribution uniformity by calculating the standard deviation of the distribution of all data points within the sub-region across each feature dimension. A smaller standard deviation indicates a more uniform distribution. The distribution uniformity was normalized to a range of 0 to 1. The calculated standard deviation was normalized to obtain the distribution uniformity index for each sub-region. Boundary sharpness was calculated by statistically analyzing the proportion of data points belonging to the sub-region boundary; a higher proportion indicates a sharper boundary. Simultaneously, considering the characteristic differences between adjacent sub-regions, the degree of feature abrupt change at the boundary was calculated; a higher degree of abrupt change indicates a sharper boundary. The final boundary sharpness index was obtained by combining these two results and normalized to a range of 0 to 1. The calculated cluster density, distribution uniformity, and boundary sharpness indices were validated. A threshold of 0.02 was set for the index calculation error; a value was considered valid when the error between the calculated value and the manually verified value was less than 0.02. For invalid indices, the calculation method was readjusted, and the calculation process was repeated to finally obtain the quantitative index of data clustering characteristics.

[0104] Step 5.3: Quantify the data aggregation characteristics using quantitative indicators, and calculate and generate corresponding task configuration parameters for each sub-region. The task configuration parameters include the proportion of computing resources allocated, task execution priority, and data processing granularity, forming a set of task configuration parameters. Specifically, this includes: establishing a mapping rule library between the data aggregation characteristics quantitative indicators and task configuration parameters, and clarifying the parameter configuration standards corresponding to different indicator value ranges. For example, when the data cluster density is higher than 0.8, the computing resource allocation ratio is set to 30%; when the density is between 0.5 and 0.8, the allocation ratio is set to 20%; and when the density is lower than 0.5, the allocation ratio is set to 10%. The data processing granularity is adjusted based on the distribution uniformity index: when the distribution uniformity is higher than 0.8, a fine-grained processing method is used, dividing the data points into processing units of 10; when the uniformity is between 0.4 and 0.8, a medium-grained processing method is used, dividing the data points into groups of 20; and when the uniformity is lower than 0.4, a coarse-grained processing method is used, dividing the data points into groups of 50. The task execution priority is determined by combining boundary clarity and cluster density: sub-regions with boundary clarity higher than 0.9 and cluster density higher than 0.7 have a task execution priority of level one; sub-regions with boundary clarity between 0.7 and 0.9 or cluster density between 0.5 have a priority of level one. For values ​​between 0.7 and 0.7, a Level 2 priority is assigned; otherwise, a Level 3 priority is assigned, with Level 1 having the highest priority and consuming computing resources first. For each sub-region, quantified indicators are used based on its data aggregation characteristics. The configuration standards in the mapping rule base are called to calculate and generate the corresponding computing resource allocation ratio, task execution priority, and data processing granularity parameters, forming the task configuration parameters for a single sub-region. The task configuration parameters for all sub-regions are summarized to form a task configuration parameter set. The rationality of the parameter set is verified by checking whether the sum of the computing resource allocation ratios is 100%, whether the priority division conforms to the importance of the sub-region data, and whether the processing granularity is adapted to the data distribution. The parameter adaptability threshold is set to 95. When the adaptability reaches 95 or above, the parameter set is considered valid. For invalid parameters, the configuration standards are readjusted, and the parameter set is generated again.

[0105] Step 5.4 involves using a set of task configuration parameters to differentiate and adapt the original computation task logic of each privacy computing node, ensuring that the computation task logic of each node matches the data characteristics of its respective sub-region. This results in a set of adapted and adjusted task logics. Specifically, this includes: collecting the original computation task logic of each privacy computing node, identifying the core modules included in the logic (such as data reading, computation processing, and result output modules), clarifying the function, input / output format, and runtime dependencies of each module, and forming a list of original task logics; establishing a correspondence between task configuration parameters and task logic adjustment rules. For example, for nodes in sub-regions with high computational resource allocation ratios, a parallel processing branch is added to the computation processing module to improve computational efficiency; for tasks with high execution priority, a priority execution flag is set in the task scheduling module to ensure priority use of node resources; for data processing with fine granularity, a data sharding and subdivision function is added to the data reading module; and combining the data characteristics of the sub-regions corresponding to each privacy computing node with task configuration parameters, the original computation task logic of the nodes is differentiated and adjusted accordingly. For example, when processing nodes with high-density clustering data in sub-regions of respiratory diseases, the computational processing module is adjusted to multi-threaded parallel processing, allocating 30% of the node's computational resources, setting the task execution priority to level one, and splitting the data processing granularity into groups of 10 data points. Adaptability testing is performed on the adjusted task logic, simulating input of data points within the sub-region to verify whether the logic can correctly execute the computational task, whether the accuracy of the calculation results meets the standard, setting the task execution accuracy threshold to 99%, whether resource usage conforms to the configuration ratio, and whether the execution efficiency is improved by more than 10% compared to the original logic. For task logic that fails the test, the matching of the adjustment rules and parameters is re-examined, the logic module design is optimized, and testing is conducted again. All successfully tested adjusted task logics are summarized to obtain a set of adapted adjusted task logics.

[0106] Step 5.5: Based on the adapted and adjusted task logic set, generate a coordination instruction sequence and task payload for each privacy computing node. Specifically, this includes: organizing the adapted and adjusted task logic set, clarifying the task logic content, required operation steps, and task configuration parameters for each privacy computing node, and converting this information into node-recognizable instruction elements, including parameter configuration instructions, task start instructions, data reading instructions, and status feedback instructions; sorting the instruction elements according to the order of task execution to generate a coordination instruction sequence; each instruction includes an instruction identifier, execution node identifier, execution time requirements, parameter information, and error handling methods. For example, parameter configuration instructions must clearly inform the node of the computing resource allocation ratio and data processing granularity, and status feedback instructions require the node to provide task execution progress feedback every 10 seconds; collecting standardized multidimensional data points in each sub-region, and combining this with the processing requirements of the task logic to process the data... The process involves packaging the data to form the core data portion of the task payload. Simultaneously, auxiliary information such as rules and thresholds required for task processing are incorporated into the payload, along with supplementary environment configuration instructions needed for node execution, thus completing the task payload construction. Coordination instruction sequences are associated with the task payloads according to their correspondence, ensuring precise matching between the instruction sequence and the corresponding task payload for each node. The generated coordination instruction sequences undergo integrity verification to check if they cover the entire task execution process, with an instruction coverage threshold set to 100%. The task payloads are then verified for data integrity and format correctness, with a data integrity threshold set to 99.9%, ensuring no missing data and that the format meets node reading requirements. The verified coordination instruction sequences and task payloads are categorized and stored, grouped according to the privacy computing node's identifier, forming a unique instruction and payload combination for each node. Finally, coordination instruction sequences and task payloads for each privacy computing node are generated.

[0107] In this embodiment of the invention, by mapping multidimensional data points to corresponding sub-regions according to their feature attributes based on the sub-region division results to clarify the data distribution status, and calculating the data aggregation density, distribution uniformity, and boundary clarity of each sub-region to obtain quantitative indicators, and generating task configuration parameters including the allocation ratio of computing resources, task execution priority, and data processing granularity based on the quantitative indicators, and then using the parameters to differentiate and adapt the original computing task logic of each privacy computing node, and finally generating a coordinated instruction sequence and task payload, this technical means effectively overcomes the technical problems of existing task configuration parameters lacking objective data support, being highly subjective in setting, mismatch between computing task logic and data distribution characteristics leading to wasted computing resources and low execution efficiency, and the inability to leverage the computing advantages of nodes due to the lack of differentiated design of heterogeneous node task logic. This achieves the goal of providing scientific quantitative basis for task configuration parameters, realizing accurate matching between computing task logic and sub-region data characteristics, and improving the targeting and efficiency of heterogeneous node task execution.

[0108] In a preferred embodiment of the present invention, step 6 above may include:

[0109] Step 6.1: Based on the coordinated instruction sequence and task payload data packets, construct a unified scheduling framework to form centralized management data for instruction and payload flows. Specifically, this includes: identifying the core attributes of the coordinated instruction sequence and task payload data packets for all privacy-preserving computing nodes, clarifying key information such as instruction type, execution requirements, associated payload identifiers, target node identifiers, payload data size, and data transmission priority, forming a list of instruction and payload attributes; designing the core architecture of the unified scheduling framework, dividing it into instruction management, payload management, timing orchestration, routing planning, distribution and execution, and status monitoring functional areas, clarifying the responsibilities and data flow paths of each functional area, ensuring the framework covers the entire process of instruction and payload management and execution; and building the framework's software support system, deploying a distributed scheduling engine, data flow middleware, and status storage service, selecting high-concurrency processing components to ensure the framework can handle multi-node instructions and payloads. The framework's parallel processing capability is optimized, with a maximum concurrent processing capacity of 2000 instructions / second to prevent service lag under high load. A centralized data management system is built, and the core display units of the view are designed, including an instruction flow status unit, a payload flow progress unit, a node association unit, and an anomaly alarm unit. The instruction flow status unit displays the real-time status of each instruction (pending orchestration, orchestrated, pending distribution, and distributed). The payload flow progress unit displays the transmission progress and integrity of payload data packets. The node association unit visually presents the correspondence between instructions, payloads, and target nodes. The view's retrieval and statistical functions are developed, supporting the retrieval of relevant data by node identifier, instruction type, and time range, and automatically calculating the instruction processing volume, payload distribution volume, and success / failure ratio per unit time. Joint testing of the framework and view is conducted to verify the framework's scheduling response time and the real-time accuracy of the view data. After successful testing, the unified scheduling framework and centralized data management system are completed.

[0110] Step 6.2, based on centralized management data, the coordinated instruction flow is time-series orchestrated according to task dependencies and execution priorities to obtain an instruction execution time sequence table. Specifically, this includes: extracting core information from all coordinated instruction sequences based on centralized management data, including instruction identifiers, target nodes, execution priorities, task dependency descriptions, and estimated execution durations; constructing an instruction information matrix to provide a data foundation for time-series orchestration; analyzing the task dependencies between instructions; identifying instruction combinations with prerequisite execution requirements by traversing the instruction information matrix, such as data reading instructions taking precedence over computation processing instructions, and feature extraction instructions requiring completion before joint analysis instructions can be initiated, forming an instruction dependency graph and clarifying the execution order constraints of each instruction; and defining the collaborative orchestration logic based on priority and dependencies, with first-priority instructions orchestrated first, second-priority instructions second, and third-priority instructions last. Instructions with dependencies are then... After the preceding instructions are arranged, the subsequent instructions are arranged. Instructions of the same priority with no dependencies can be arranged for parallel execution to improve processing efficiency. Instructions are sorted according to the arrangement rules, and each instruction is assigned a unique execution order number and a specific execution time window is set. For example, a 10-second execution window is assigned to first-priority instructions, 15 seconds to second-priority instructions, and 20 seconds to third-priority instructions. The preceding dependent instructions of each instruction are identified, and an initial instruction execution timing table is generated. The rationality and feasibility of the timing table are verified, and problems such as priority conflicts, missing dependencies, or overlapping execution time windows are checked. The timing arrangement conflict rate is calculated and is required to be less than 1. At the same time, the execution flow of the timing table is simulated to verify the continuity and time rationality of instruction execution. The timing rationality compliance rate is required to be no less than 98%. For timing tables with problems, the arrangement rules and time windows are readjusted, and after verification again, the final instruction execution timing table is obtained.

[0111] Step 6.3: Based on the instruction execution timing table, perform routing planning for the task payload flow, determine the target node path and distribution strategy for each payload data packet, and generate a routing planning scheme. Specifically, this includes: combining the instruction execution timing table and centralized management data to extract key information of the task payload flow, including payload identifier, corresponding instruction identifier, target node identifier, payload data volume, data transmission priority, and transmission time limit requirements; simultaneously, calling the global routing table of the central node to obtain the effective connection path information for each target node; formulating routing selection criteria, clarifying the core requirements that the routing plan must meet; prioritizing the path with the lowest transmission latency; when multiple paths have similar latency, selecting the path with bandwidth utilization below 60%; for large payloads with data volume greater than 1GB, prioritizing paths that support fragmented transmission to ensure transmission stability; designing a task payload distribution strategy, defining differentiated strategies based on payload data volume and transmission time limits: payloads with data volume less than 100MB and urgent transmission time limits adopt a real-time distribution strategy, immediately initiating transmission; payloads with data volume between 100MB and 1GB... The payloads are distributed using a near real-time strategy, with transmission starting 30 seconds before the corresponding instruction orchestration execution time. Payloads larger than 1GB employ a batch fragmentation strategy, splitting the payload into 100MB sub-payloads, which are transmitted sequentially with a 500ms interval between fragments to prevent network congestion. A routing planning process is executed, matching each task payload data packet with a suitable target node path, determining the corresponding distribution strategy, clarifying the sub-payload splitting rules, transmission start time, and transmission completion time limit, and generating an initial routing plan containing payload identifiers, target nodes, path information, distribution strategy, and transmission time nodes. The effectiveness of the routing plan is verified by simulating different payload transmission scenarios, testing path transmission stability, distribution strategy execution efficiency, and transmission completion time limit compliance. The path transmission accuracy threshold is set at 99.8%, the distribution strategy adaptation rate threshold at 99%, and the transmission time limit compliance rate threshold at 98.5%. For any non-compliant parts, path selection or distribution strategy parameters are adjusted, and the final routing plan is obtained after optimization.

[0112] Step 6.4: According to the routing plan, the coordination instructions and task payloads are distributed to the corresponding privacy computing nodes through the transparent data flow forwarding mechanism of the central coordination node, forming a distribution execution state. Specifically, this includes: calling the transparent data flow forwarding mechanism of the central coordination node to complete the adaptation configuration between the mechanism and the routing plan, ensuring that the forwarding mechanism can identify the path information and distribution strategy in the plan, and simultaneously initiating security verification of the forwarding channel to confirm that the channel's encrypted transmission function is normal, avoiding privacy leaks during data transmission; according to the timing requirements of the routing plan, the distribution process of coordination instructions and task payloads is initiated in batches: first, the payloads and instructions corresponding to the pre-dependent instructions are distributed, then the associated instructions and payloads are distributed; payloads and instructions using a real-time distribution strategy are prioritized for distribution, and batch-fragmented payloads are distributed sequentially according to the splitting order, while recording the distribution start time and current transmission progress of each instruction and each payload; and real-time collection of key data during the distribution process. Key data includes instruction distribution success rate, payload transmission completion rate, transmission rate, transmission delay, and abnormal information such as transmission interruption and target node non-response. The distribution status is divided into four types: pending distribution, distribution in progress, distribution successful, and distribution failed. For instructions and payloads that fail to be distributed, a retry mechanism is automatically triggered, with the number of retries set to 3, and each retrieval interval being 10 seconds. If the retry still fails, it is marked as a distribution failure and the reason for failure is recorded. Distribution result verification is carried out. For instructions and payloads marked as successfully distributed, a reception confirmation request is sent to the target privacy computing node to verify whether the instructions and payloads are completely received and whether the format is correct. The reception confirmation accuracy threshold is set to 99.9%. For instructions and payloads with abnormal reception confirmation, the distribution process is restarted. All instruction and payload distribution status data are summarized, classified and organized by target node, and a distribution execution status report containing distribution status, transmission details, and abnormal records is generated, thus completing the formation of the distribution execution status.

[0113] Step 6.5: Based on the distributed execution status, monitor the local computation execution process of each privacy computing node, collect computation status information fed back by the nodes, and centrally coordinate and dynamically adjust abnormal execution states. This ensures that the central coordination node is only responsible for instruction scheduling and status monitoring without participating in actual computation. Ultimately, this achieves a collaborative architecture where computation tasks are executed on local nodes while the management process is centralized at the central node. Specifically, this includes: establishing a monitoring indicator system for the computation execution process, identifying the core indicators to be monitored, including CPU utilization, memory usage, task execution progress, instruction execution success rate, data processing throughput, and abnormal errors for each privacy computing node. The system sets a monitoring cycle of 5 seconds to ensure real-time capture of node execution status changes. It collects computing status information from each privacy computing node through interconnection channels, compares the collected information with the monitoring indicator system, and dynamically updates node execution status in centralized management data. Nodes with CPU utilization exceeding 85% and memory usage exceeding 80% are marked as resource-scarce. Nodes with task execution progress more than 50% below the timing table requirement and instruction execution success rate below 95% are marked as execution-abnormal. Specific adjustment strategies are defined for different abnormality types: in resource-scarce states, a dynamic computing resource allocation process is initiated to reduce... If the resource allocation ratio for low-priority tasks on this node is reduced from 10% to 5%, the released resources will be reallocated to high-priority tasks. In case of execution anomalies, a status check command is first sent to the node to obtain the detailed cause of the anomaly. If the anomaly is a command format issue, the corrected command is redistributed; if it is a data loss issue, the corresponding task payload is redistributed; if it is a node hardware failure, a backup node is activated to take over the task. Strictly adhering to the principle that the central coordinating node does not participate in actual computation, all adjustment operations are completed through coordination commands. The central node is only responsible for command generation, distribution, and status collection and judgment, and does not touch any node's original data or computation. The system calculates process data and records all abnormal adjustments, including adjustment time, strategy, and effect, forming an exception handling log. It verifies the effectiveness of the collaborative architecture by monitoring data to ensure that the local execution rate of computational tasks reaches 100%, the centralization of management at the central node achieves 100% coverage of instruction scheduling and status monitoring, the task execution timing compliance rate is no less than 98%, and the improvement in computational resource utilization is no less than 20%. For those that do not meet the standards, the system optimizes monitoring indicator thresholds or abnormal adjustment strategies, ultimately achieving a collaborative architecture where computational tasks are executed locally while management is centralized at the central node.

[0114] In this embodiment of the invention, by constructing a unified scheduling framework based on the coordination instruction sequence and task payload data packets and forming centralized management data, the timing of the coordination instruction stream is arranged according to task dependencies and execution priorities based on this view. The target node paths and distribution strategies of the task payload stream are planned. Instructions and payloads are distributed through the transparent forwarding mechanism of the central coordination node. At the same time, the local computing execution process of each node is monitored and abnormal states are centrally coordinated and dynamically adjusted. The central node is only responsible for instruction scheduling and status monitoring and does not participate in actual computing. This technical approach effectively overcomes the technical problems of existing instructions and payloads lacking a unified scheduling mechanism, resulting in chaotic task execution timing, unreasonable payload distribution paths, poor task execution stability due to the lack of real-time monitoring and abnormal handling mechanisms for node computing status, and the risk of data privacy leakage due to unclear boundaries between centralized management and local computing. This achieves the technical effect of orderly scheduling and accurate distribution of instruction streams and payload streams, ensuring the stability and reliability of multi-terminal task execution, clarifying the responsibilities of the central node and privacy computing nodes, and ultimately realizing efficient collaboration between computing localization and centralized management. This maximizes the protection of data privacy and security while improving the overall efficiency of multi-terminal data privacy collaborative processing.

[0115] like Figure 2 As shown, embodiments of the present invention also provide a multi-terminal data privacy collaborative processing system based on a central network node, including:

[0116] The acquisition module is used to deploy multiple heterogeneous privacy computing nodes with different privacy computing protocols, network communication protocols, and upper-layer application designs.

[0117] The adjustment module is used to set up a central coordination node to address the need for interconnection and interoperability of heterogeneous privacy computing nodes. The central coordination node has protocol compatibility capabilities.

[0118] The connection module is used to enable each heterogeneous privacy computing node to establish a network connection with the central node through the protocol compatibility of the central node, and to establish an indirect interconnection channel between all privacy computing nodes through the central node.

[0119] The partitioning module is used to collect source data from each node based on the established interconnection channels and transform it into a multi-dimensional data point set. By performing cluster boundary analysis on the multi-dimensional data point set, the cluster boundary information of the data point set is obtained. Based on the cluster boundary information, a reference plane is constructed, the base region is defined, and interrelated sub-regions are divided according to the data distribution and cluster boundary information to form the sub-region partitioning result.

[0120] The adjustment module is used to map multidimensional data points to corresponding sub-regions based on the sub-region division results, obtain task configuration parameters based on the aggregation characteristics of data points in each sub-region, adapt and adjust the computing task logic of each privacy computing node based on the task configuration parameters, and obtain corresponding coordination instructions and task payloads by the central coordination node based on the adjusted task logic.

[0121] The processing module is used to coordinate the instruction stream and task payload stream for unified scheduling and distribute routes to the corresponding privacy computing nodes, so as to schedule each node to execute computing tasks locally, thereby achieving collaborative processing of computing localization and centralized management.

[0122] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for multi-end data privacy collaborative processing based on a central networking node, characterized in that, The method comprises: Deploying a plurality of heterogeneous privacy computing nodes of different privacy computing protocols, network communication protocols and upper layer application designs; Setting a central coordination node for the interconnection and intercommunication needs of the heterogeneous privacy computing nodes, and the central coordination node has protocol compatibility capability; Through the protocol compatibility capability of the central node, each heterogeneous privacy computing node is connected to the central node once, and the indirect interconnection and intercommunication channel between all privacy computing nodes is established through the central node; Based on the established interconnection and intercommunication channel, the central node collects source data of each node and converts it into a multi-dimensional data point set; through clustering boundary analysis of the multi-dimensional data point set, data point clustering boundary information is obtained; based on the clustering boundary information, a reference plane is constructed, a base area is defined, and sub-areas related to each other are divided according to data distribution and clustering boundary information, forming a sub-area division result; Through the sub-area division result, the multi-dimensional data points are mapped to the corresponding sub-areas, and the task configuration parameters are obtained according to the aggregation characteristics of the data points in each sub-area; through the task configuration parameters, the computing task logic of each privacy computing node is adapted and adjusted, and based on the adjusted task logic, the central coordination node obtains the corresponding coordination instructions and task load; Through unified scheduling of the coordination instruction stream and the task load stream, and distributing the route to the corresponding privacy computing node, the local execution of the computing task of each node is scheduled, realizing the collaborative processing of computing localization and management centralization.

2. The hub-and-spoke based multi-party data privacy co-processing method of claim 1, wherein, Deploying a plurality of heterogeneous privacy computing nodes of different privacy computing protocols, network communication protocols and upper layer application designs, comprising: Performing heterogeneous characteristic identification analysis on each privacy computing node to be accessed to obtain analysis results of differences in privacy computing protocol types, network communication protocol types and upper layer application design architecture; According to the heterogeneous characteristic identification analysis result, configure the privacy computing protocol compatibility capability, so that the privacy computing protocol compatibility capability supports the conversion processing of RSA+Hash, national secret+Hash, ECDH, OPRF and VOLE multiple privacy computing protocols; Based on the configuration result of the privacy computing protocol compatibility capability, establish the network communication protocol adaptation capability, so that the network communication protocol adaptation capability processes the protocol conversion of single network communication request-pull mode, long connection mode and gRPC protocol, and obtains the establishment result of the network communication protocol adaptation capability; Through the establishment result of the network communication protocol adaptation capability, generate the application interface standardization capability, so that the application interface standardization capability unifies the application interface specifications of different heterogeneous privacy computing platforms, realizes the standardized mapping of heterogeneous application design styles, and obtains the generation result of the application interface standardization capability; According to the generation result of the application interface standardization capability, integrate the privacy computing protocol compatibility capability, the network communication protocol adaptation capability and the application interface standardization capability, and complete the unified deployment of the plurality of heterogeneous privacy computing nodes.

3. The hub-and-spoke based multi-party data privacy co-processing method of claim 2, wherein, Setting a central coordination node for the interconnection and intercommunication needs of the heterogeneous privacy computing nodes, and the central coordination node has protocol compatibility capability, comprising: Based on the networking needs of the interconnection and intercommunication of the heterogeneous privacy computing nodes, build the basic framework of the central coordination node; The infrastructure of the central coordination node is configured to configure a network protocol compatible layer, so that the network protocol compatible layer identifies and converts communication data in a single network communication request-pull mode, a long connection mode and a gRPC protocol, and obtains a configuration result of the network protocol compatible layer; According to the configuration result of the network protocol compatible layer, a privacy computing protocol conversion capability is constructed, so that the constructed privacy computing protocol conversion capability supports real-time conversion and adaptation of multiple privacy computing protocols such as RSA+Hash, national secret+Hash, ECDH, OPRF and VOLE, and obtains a construction result of the privacy computing protocol conversion capability; According to the construction result of the privacy computing protocol conversion capability, an application layer interface standardization mechanism is established, so that the application layer interface standardization mechanism uniformly maps the application interface specifications of different heterogeneous privacy computing platforms into a standard interface format recognizable by the central coordination node, and obtains an establishment result of the application layer interface standardization mechanism; Through the establishment result of the application layer interface standardization mechanism, the network protocol compatible layer, the privacy computing protocol conversion capability and the application layer interface standardization mechanism are integrated to form the complete protocol compatible capability of the central coordination node.

4. The hub-and-spoke based multi-party data privacy co-processing method of claim 3, wherein, Through the protocol compatible capability of the central node, each heterogeneous privacy computing node is connected to the central node once, and an indirect interconnection channel between all privacy computing nodes is established through the central node, including: Through the complete protocol compatible capability of the central coordination node, the networking service of the central node is initialized to provide a unified access endpoint for the heterogeneous privacy computing nodes, and a networking service initialization state is obtained; According to the networking service initialization state, the networking connection request of each heterogeneous privacy computing node is received, and the network protocol recognition and protocol conversion processing are performed on each networking connection request to obtain a converted standardized networking request; Based on the converted standardized networking request, a single networking connection between each heterogeneous privacy computing node and the central node is established, and a node and central connection mapping relationship table is obtained; According to the node and central connection mapping relationship table, a global routing table of the central node is constructed, so that the global routing table forwards data flow from any source node to the corresponding target node through the central node according to the target node identifier, and obtains a routing and forwarding capability of the central node; Through the routing and forwarding capability of the central node, the data flow transparent forwarding mechanism of the central node is activated, so that the central node only performs data routing distribution without participating in actual calculation, and an indirect interconnection channel between all privacy computing nodes is established.

5. The hubbing node based multi-end data privacy co-processing method according to claim 4, characterized in that, Based on the established interconnection channel, the central node collects source data of each node and converts it into a multi-dimensional data point set; through clustering boundary analysis on the multi-dimensional data point set, data point clustering boundary information is obtained; Based on the clustering boundary information, a reference plane is constructed, a base region is defined, and sub-regions are divided according to data distribution and clustering boundary information, forming a sub-region division result, including: Source data feature information is collected from each privacy computing node through the interconnection channel and converted into a standardized multi-dimensional data point set; According to the standardized representation of the multi-dimensional data point set, a point set clustering boundary extraction algorithm is executed to analyze the density distribution and identify the boundary of the multi-dimensional data points, and the data point set clustering boundary information is obtained; According to the data point set clustering boundary information, a data distribution reference plane is constructed, so that the data distribution reference plane accurately reflects the relative position relationship of each data cluster in the multi-dimensional space, and a reference plane structure with geometric characteristics is obtained; Based on the reference plane structure with geometric characteristics, a data processing base area is defined, so that the data processing base area covers the distribution range of all data clusters and retains the correlation between data, and a complete base area definition is obtained; Through the complete base area definition, combined with the clustering boundary information and the data distribution density characteristics, the base area is divided into multiple interrelated sub-areas, each sub-area corresponds to a specific data distribution characteristic and a calculation task type, and a sub-area division result is formed.

6. The hubbing node based multi-end data privacy co-processing method of claim 5, wherein, Through the sub-area division result, the multi-dimensional data points are mapped to the corresponding sub-areas, and the task configuration parameters are obtained according to the aggregation characteristics of the data points in each sub-area; through the task configuration parameters, the calculation task logic of each privacy calculation node is adapted and adjusted, and based on the adjusted task logic, the corresponding coordination instructions and task loads are obtained by the central coordination node, including: According to the sub-area division result, the multi-dimensional data points are mapped to the corresponding sub-areas according to the data feature attributes, and the data point distribution state in each sub-area is obtained; According to the data point distribution state in each sub-area, the data aggregation density, distribution uniformity and boundary clarity characteristics of each sub-area are calculated, and the data aggregation characteristic quantitative index is obtained; Through the data aggregation characteristic quantitative index, the corresponding task configuration parameters are calculated for each sub-area, including the calculation resource allocation ratio, the task execution priority and the data processing granularity, forming a task configuration parameter set; Through the task configuration parameter set, the original calculation task logic of each privacy calculation node is differentially adapted and adjusted, so that the calculation task logic of each node matches the data characteristics of the sub-area where it is located, and an adjusted task logic set is obtained; Based on the adjusted task logic set, the coordination instruction sequence and the task load for each privacy calculation node are generated.

7. The hub-and-spoke based multi-party data privacy co-processing method of claim 6, wherein, Through unified scheduling of the coordination instruction stream and the task load stream, and routing is distributed to the corresponding privacy calculation node, to schedule each node to execute the calculation task locally, to realize the collaborative processing of calculation localization and management centralization, including: According to the coordination instruction sequence and the task load data packet, a unified scheduling framework is constructed, forming a centralized management data of instruction stream and load stream; Based on the centralized management data, the coordination instruction stream is time-sequenced according to the task dependency relationship and the execution priority, and an instruction execution time sequence table is obtained; According to the instruction execution time sequence table, the task load stream is routed and planned, the target node path and the distribution strategy of each load data packet are determined, and a routing planning scheme is generated; According to the routing planning scheme, through the data stream transparent forwarding mechanism of the central coordination node, the coordination instructions and the task loads are distributed to the corresponding privacy calculation nodes, forming a distribution execution state; Based on the distribution execution state, the local computing execution process of each privacy computing node is monitored, the computing state information fed back by the nodes is collected, the abnormal execution state is centrally coordinated and dynamically adjusted, the central coordination node is only responsible for instruction scheduling and state monitoring without participating in actual computing, and finally the cooperative architecture that the computing task is executed on the local node and the management process is centralized on the central node is realized.

8. A multi-end data privacy co-processing system based on a central hub node, the system implementing the method of any one of claims 1 to 7, characterized in that, The method comprises the steps of: An acquisition module is configured to deploy a plurality of heterogeneous privacy computing nodes designed based on different privacy computing protocols, network communication protocols, and upper-layer applications; An adjustment module is configured to set a central coordination node based on the demand for interconnection and intercommunication of the heterogeneous privacy computing nodes, wherein the central coordination node has protocol compatibility capability; A connection module is configured to enable each of the heterogeneous privacy computing nodes to be connected to the central coordination node once through the protocol compatibility capability of the central coordination node, and to establish an indirect interconnection and intercommunication channel between all the privacy computing nodes through the central coordination node; A division module is configured to collect source data of each node by the central coordination node based on the established interconnection and intercommunication channel, and to convert the source data into a multi-dimensional data point set; A clustering boundary analysis is performed on the multi-dimensional data point set to obtain data point clustering boundary information; a reference plane is constructed based on the clustering boundary information, a base region is defined, and a plurality of sub-regions are divided based on data distribution and clustering boundary information, thereby forming a sub-region division result; An adjustment module is configured to map the multi-dimensional data points to the corresponding sub-regions based on the sub-region division result, obtain task configuration parameters based on the aggregation characteristics of the data points in each sub-region, adaptively adjust the computing task logic of each privacy computing node based on the task configuration parameters, and obtain corresponding coordination instructions and task loads based on the adjusted task logic by the central coordination node; A processing module is configured to uniformly schedule the coordination instruction stream and the task load stream, and to distribute routes to the corresponding privacy computing nodes to schedule the local execution of the computing task by each node, thereby realizing cooperative processing of computing localization and management centralization.

9. A computing device, comprising: The method comprises the steps of: One or more processors; A storage device is configured to store one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a program, which is executed by a processor to implement the method according to any one of claims 1 to 7.

Images