Distributed credential and credential privacy computing system
The distributed domestically developed privacy computing system, through its layered architecture, cloud-edge-device collaboration, and communication management modules, solves the problems of data security and computational stability in privacy computing systems, achieving data privacy protection and efficient and reliable execution of computing tasks.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-24
- Publication Date
- 2026-03-10
AI Technical Summary
Existing privacy computing systems lack data privacy and security, pose a risk of personal information leakage, and have poor stability and reliability in computing tasks.
The system employs a layered architecture design, including a security isolation module, a cloud-edge-device collaborative operation framework, and a communication management module. Combined with a white-box testing verification platform, it achieves isolation and collaboration between the data layer, computing layer, and application layer. It ensures the self-healing of the communication link through heartbeat monitoring and master/slave socket switching strategies, and performs intelligent testing and verification in a simulation environment.
While ensuring data privacy and security, it can reliably and stably complete complex intelligent tasks, eliminate the risk of large-scale data leakage, reduce computing latency and network costs, and achieve self-verification and continuous optimization.
Smart Images
Figure CN121644129A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of information technology innovation security, and specifically relates to a distributed information technology innovation privacy computing system. Background Technology
[0002] With the rapid development of intelligent operation and maintenance, cloud computing, and big data computing, the data demands of various industries are rising sharply, and the increasing accumulation of central data is placing higher and higher demands on the deployment and delivery of integrated machines. Domestically developed databases are widely used in various industries, covering finance, telecommunications, government, manufacturing, healthcare, and many other fields. With the continuous development of information technology and the advancement of the national information security strategy, privacy-preserving computing systems based on domestically developed desktop applications are computing-aided tools that integrate artificial intelligence technology. They can intelligently analyze and process input data through algorithms such as machine learning and deep learning, thereby providing accurate calculation results and valuable decision-making suggestions.
[0003] However, privacy-preserving computing systems require access to users' personal information and data to provide personalized services, which increases the risk of privacy breaches. If security measures are inadequate, users' personal information may be misused or leaked, leading to violations of their privacy. Summary of the Invention
[0004] This invention addresses the shortcomings of existing technologies by providing a distributed domestically developed privacy computing system that, while ensuring data privacy and security, fully utilizes distributed computing power to stably and reliably complete complex intelligent tasks and can be continuously optimized.
[0005] This invention provides the following technical solution: A distributed domestically developed privacy computing system includes: The security isolation module adopts a layered architecture design, dividing the privacy computing system into a data layer, a computing layer, and an application layer. The data layer is used to encrypt and store the original privacy data, the computing layer is used to execute privacy computing tasks, and the application layer is used to provide a user interface. The layers are isolated and communicate with each other through API interfaces and encryption protocols. The cloud-edge-device collaborative operation framework is used to support the operation of the privacy computing system. Through the collaboration of cloud center management nodes, edge nodes and domestically developed desktop clients, it jointly realizes the distributed execution of privacy computing tasks. The communication management module monitors the connection health status in real time through heartbeat monitoring and adopts a primary / backup socket switching strategy to achieve self-healing of communication link failures. The verification platform, based on white-box testing methods, tests and verifies intelligent computing algorithms, communication logic, and system processes that are about to be executed or are currently being executed in a simulation environment.
[0006] Optionally, the security isolation module further includes: a domestic IT innovation database, which provides static data encryption for the data layer, data protection during privacy computing and encryption before transmission for the computing layer, and transmission encryption and authentication for the application layer.
[0007] Optionally, the cloud center management node performs global task scheduling and resource allocation through a task orchestrator, and uses a model aggregation server to aggregate global models of federated learning during the privacy computing process; The edge node connects to the cloud center management node and the domestically developed desktop client through an edge computing gateway, and initially aggregates data or models from multiple desktop clients; The aforementioned domestically developed desktop client executes computing tasks in the desktop environment through a local task executor and maintains state synchronization with the cloud center management node and edge nodes through a state synchronizer.
[0008] Optionally, the communication management module includes a connection management service component, a fault detection and switching engine, and a domestically developed socket API; The fault detection and switching engine uses a heartbeat monitoring mechanism to periodically check the health status of the primary and backup sockets, and makes a decision to switch between the primary and backup sockets when an abnormality is detected. The domestic socket API executes primary / standby socket failover decisions. The connection management service component integrates a primary / secondary socket pairing table and a scheduler. The connection management service component can record the primary / secondary relationship and health status of all connections in real time. When a primary / secondary socket pair needs to be created, the scheduler will schedule the primary and secondary sockets to different physical nodes or different memory spaces.
[0009] Optionally, when an abnormal health status occurs, anti-ping-pong handover logic is introduced. When the heartbeat detection of the primary socket or the backup socket fails for the first time, the link quality of the primary socket or the backup socket is continuously monitored within a set time, and a decision to cancel the handover or a handover decision is issued.
[0010] Optionally, the verification platform, based on white-box testing methods, tests and verifies the intelligent computing algorithms, communication logic, and system processes that are about to be executed or are being executed in a simulation environment, specifically as follows: In the simulation environment, the Fireworks algorithm is used independently to search for input data or training data that cause boundary condition errors or numerical instability in the privacy computing model, to search for communication network sequences that cause master-slave switching failures or ping-pong switching, and to search for node failure sequences that cause deadlocks or inconsistent states in the privacy computing system, in order to optimize the intelligent computing algorithm and discover vulnerabilities in communication logic and system processes.
[0011] Optionally, maximize the fitness function The Fireworks algorithm continuously generates data that causes boundary condition errors or data instability in the privacy computing model, including batches of training data for the privacy computing model and hyperparameters of the privacy computing model. ; in, , and These represent the number of times the privacy-preserving computation model contains non-NaN values, the number of times gradient explosion occurs, and the magnitude of the abrupt change in the loss function value, respectively. , and They are respectively , and The corresponding weights.
[0012] Optionally, maximize the fitness function The fireworks algorithm is used to search for communication network sequences that have failed primary / backup switching or triggered ping-pong switching, including latency, packet loss rate and fault duration. ; in, , and These represent the number of communication network sequence handover failures, the number of ping-pong handovers per unit time, and the average fault recovery time, respectively. , and They are respectively , and The corresponding weights.
[0013] Optionally, maximize the fitness function The Fireworks algorithm is used to search for the timing of node failures that cause deadlocks or inconsistencies in the privacy computing system, including the faulty node, the type of failure, and the time of failure. ; in, , and These represent the number of times the privacy-preserving computation system experienced a complete deadlock, the number of nodes with inconsistent states, and the percentage decrease in the integrity rate of privacy-preserving computation tasks. , and They are respectively , and The corresponding weights.
[0014] Compared with the prior art, the beneficial effects of the present invention are: The security isolation module, cloud-edge-device collaborative operation framework, communication management module, and verification platform of this invention, while ensuring data privacy and security, fully utilize distributed computing power to stably and reliably complete complex intelligent tasks, and can self-verify and continuously optimize. Specifically, the security isolation module adopts a layered design, ensuring that sensitive data is always kept locally, eliminating the risk of large-scale data leakage at the source. The cloud-edge-device collaborative operation framework significantly reduces computing latency, optimizes network bandwidth costs, and aggregates distributed computing power to achieve collaborative and efficient utilization of computing power. The communication management module can automatically and seamlessly switch to backup links, ensuring that the failure of a single node or network link will not lead to the failure of the entire computing task. The verification platform based on white-box testing can discover deep-seated vulnerabilities in advance before the privacy computing system goes online, enabling the system to have the ability to self-verify and continuously optimize. Attached Figure Description
[0015] Figure 1 This is a structural block diagram of the distributed domestic information technology-based privacy computing system of the present invention; Figure 2 This is a structural block diagram of the cloud-edge-device collaborative operation framework of the present invention; Figure 3 This is a structural block diagram of the communication management module of the present invention. Detailed Implementation
[0016] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the scope of protection of the present invention. It should be noted that the term "comprising" and any variations thereof in the specification, claims and the above-mentioned drawings of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or devices.
[0017] like Figure 1 As shown, a distributed domestically developed privacy computing system includes: a security isolation module, a cloud-edge-device collaborative operation framework, a communication management module, and a verification platform. Under the premise of ensuring data privacy and security, it makes full use of distributed computing power to stably and reliably complete complex intelligent tasks, and can self-verify and continuously optimize.
[0018] I. Security Isolation Module The privacy computing system adopts a layered architecture design, dividing it into a data layer, a computing layer, and an application layer.
[0019] The data layer is used to encrypt and store raw privacy data, defining how data is encrypted, stored, and accessed. The computation layer is used to perform privacy-preserving computation tasks. The application layer provides the user interface; the layers are isolated and communicate with each other through API interfaces and encryption protocols. The security isolation module ensures that data does not leave the domain. This layered architecture design ensures that sensitive data is always kept locally (on the domestically developed desktop or edge node), without needing to be uploaded to a central server, fundamentally eliminating the risk of large-scale data leaks and simplifying core computational logic.
[0020] In this embodiment, the security isolation module further includes: a domestic IT innovation database.
[0021] The key to the success of domestically developed databases lies in their compliance with national information security and data autonomy requirements. They must support domestic independent research and development and technological innovation, and be able to replace foreign products in critical information infrastructure, thereby ensuring national information security. Domestically developed databases integrate national cryptographic algorithms (SM2 / SM4 / SM9).
[0022] The domestic IT database provides static data encryption for the data layer.
[0023] Specifically, when raw privacy data is persistently stored and written to a hard drive or database, the data layer calls the Inter Milan algorithm to encrypt the data, thereby ensuring the static security of the data.
[0024] The domestically developed database provides data protection and encryption before transmission for privacy computing at the computing layer.
[0025] Specifically, during privacy-preserving computations, such as federated learning, decrypted data or intermediate computation results (such as model gradients) are processed in memory. The computation layer needs to call relevant functions of the domestically developed database to ensure that the data in memory is not illegally accessed. Furthermore, before the computation layer sends the computation results to other nodes for aggregation, it calls the domestically developed database to encrypt the data to be transmitted. The settings of the domestically developed database ensure dynamic data security.
[0026] The domestic IT database provides transmission encryption and authentication for the application layer.
[0027] Encryption during transmission: When the application layer sends a request to the computing layer or receives a result, it calls the encryption library to establish a TLS / SSL (using Chinese cryptographic algorithms) secure link to ensure that the data is encrypted when transmitted over the network.
[0028] Identity authentication: When the application layer interacts with the lower layer, it calls SM2 to perform digital signature to verify the authenticity of the identity.
[0029] II. Cloud-Edge-Device Collaborative Operation Framework It is used to support the operation of the privacy computing system. Through the collaboration of cloud center management nodes, edge nodes and domestically developed desktop clients, it can jointly realize the distributed execution of privacy computing tasks.
[0030] like Figure 2 As shown, the cloud center management node is responsible for global task orchestration, model aggregation, and resource scheduling. The cloud center management node includes a task orchestrator and a model aggregation server. The task orchestrator is used to receive computing tasks, decompose them into sub-tasks, and distribute them to edge nodes and domestically developed desktop clients. The model aggregation server is used to aggregate model parameters uploaded by various edge nodes. The specific aggregation method can refer to existing technologies.
[0031] Edge nodes, acting as an intermediate layer, undertake some computing and aggregation tasks, alleviating cloud pressure and reducing latency. Edge nodes include: edge computing gateways; these gateways connect cloud center management nodes and domestically developed desktop clients, and initially aggregate data or models from multiple desktop clients; in some other embodiments, edge nodes also include data preprocessing engines, which perform data cleaning and feature extraction at the edge.
[0032] The domestically developed desktop client executes local computing tasks and provides a user interface. It includes a local task executor for executing computing tasks in the desktop environment. Typically, the domestically developed desktop client also includes a state synchronizer to keep the state of the domestically developed desktop client synchronized with the cloud center management node and edge nodes.
[0033] The cloud-edge-device collaborative operation framework significantly optimizes memory usage, reduces communication bandwidth consumption between nodes, and shortens the overall completion time of distributed training tasks, making it highly feasible to deploy intelligent computing in resource-constrained edge environments.
[0034] III. Communication Management Module The connection health status is monitored in real time by heartbeat monitoring, and a primary / backup socket switching strategy is adopted to achieve self-healing of communication link failures.
[0035] like Figure 3 As shown, the communication management module includes a connection management service component, a fault detection and switching engine, and a domestically developed socket API; The fault detection and switching engine uses a heartbeat monitoring mechanism to periodically check the health status of the primary and backup sockets, and makes a decision to switch between the primary and backup sockets when an abnormality is detected.
[0036] The heartbeat monitoring mechanism, referencing existing technologies, specifically involves periodically sending heartbeat packets at fixed time intervals. If a response is received within the set timeout period, the current primary or backup socket is determined to be healthy. If no response is received within the set timeout period, the current primary or backup socket is determined to be in an abnormal or faulty state.
[0037] The domestic socket API executes primary / standby socket failover decisions.
[0038] The connection management service component integrates a primary / standby socket pairing table and a scheduler. This component records the primary / standby relationship and health status (normal, faulty, and switching over) of all connections in real time. The scheduler, when a primary / standby socket pair needs to be created, schedules the primary and standby sockets to different physical nodes or different memory spaces. The primary / standby socket pairing table is shown in Table 1, assigning primary and standby sockets to different memory regions, such as starting with C and D.
[0039] Table 1. Pairing table for primary and backup sockets
[0040] In this embodiment, when the health status is abnormal, anti-ping-pong handover logic is introduced. When the heartbeat detection of the primary socket is abnormal for the first time, the link quality of the primary socket is continuously monitored for a set time, and a decision to cancel the handover or a handover decision is issued. Similarly, when the health status of the backup socket is abnormal, the link quality of the backup socket is continuously monitored for a set time.
[0041] Link quality includes: latency, the degree of latency variation, packet loss rate, bandwidth or throughput. When monitoring link quality, you can set weighted scores or set multi-level thresholds.
[0042] In this embodiment, after the primary / backup socket switch is performed, the communication management module will continuously monitor the status of the original socket. If it recovers, it will be set as the backup, or set as the primary socket according to the user's choice. If the status of the original socket is continuously monitored as abnormal, a new primary / backup socket pair will be created.
[0043] The communication management module significantly enhances the reliability and robustness of the privacy computing system. The failure of a single node or process will not cause the entire computing task to fail, ensuring the stable execution of long-term privacy computing tasks.
[0044] IV. Verification Platform Based on white-box testing methods, intelligent computing algorithms, communication logic, and system processes that are about to be executed or are being executed are tested and verified in a simulation environment to efficiently discover vulnerabilities and optimize parameters, thereby ensuring the overall reliability and security of the system.
[0045] Intelligent computing algorithms: For example, testing whether the aggregation algorithm of federated learning models is correct and whether it will be biased by data from certain malicious nodes.
[0046] Communication logic: For example, testing whether the primary / backup socket switching process is rigorous and whether the anti-ping-pong mechanism is effective.
[0047] System process: For example, testing whether the entire chain of a complete computing task, from issuance and execution to aggregation, is smooth.
[0048] In this implementation, in a simulation environment, the Fireworks algorithm is used independently to search for input data or training data that cause boundary condition errors or numerical instability in the privacy computing model, to search for communication network sequences that cause master-slave switching failures or ping-pong switching, and to search for node failure sequences that cause deadlocks or inconsistent states in the privacy computing system, in order to optimize the intelligent computing algorithm and discover vulnerabilities in communication logic and system processes.
[0049] The Fireworks algorithm is used to search for input or training data that introduces boundary condition errors or numerical instability into the privacy-preserving computation model. Specifically, it maximizes the fitness function. The Fireworks algorithm continuously generates data that causes boundary condition errors or data instability in the privacy computing model, including batches of training data for the privacy computing model and hyperparameters of the privacy computing model. ; in, , and These represent the number of times the privacy-preserving computation model contains non-NaN values, the number of times gradient explosion occurs, and the magnitude of the abrupt change in the loss function value, respectively. , and They are respectively , and The corresponding weights.
[0050] The fireworks algorithm is used to search for communication network sequences that exhibit master-slave handover failures or trigger ping-pong handovers. Specifically, it involves maximizing the fitness function. The fireworks algorithm is used to search for communication network sequences that have failed primary / backup switching or triggered ping-pong switching, including latency, packet loss rate and fault duration. ; in, , and These represent the number of communication network sequence handover failures, the number of ping-pong handovers per unit time, and the average fault recovery time, respectively. , and They are respectively , and The corresponding weights.
[0051] The Fireworks algorithm is used to search for node failure sequences that cause deadlocks or state inconsistencies in the privacy-preserving computation system, specifically: maximizing the fitness function. The Fireworks algorithm is used to search for the timing of node failures that cause deadlocks or inconsistencies in the privacy computing system, including the faulty node, the type of failure, and the time of failure. ; in, , and These represent the number of times the privacy-preserving computation system experienced a complete deadlock, the number of nodes with inconsistent states, and the percentage decrease in the integrity rate of privacy-preserving computation tasks. , and They are respectively , and The corresponding weights.
[0052] The design of the verification platform can discover deep vulnerabilities and logical defects in the system under concurrency, load and fault simulation earlier and more efficiently than traditional testing methods, thereby significantly improving the system's intelligence level and decision-making accuracy before deployment.
[0053] When a user submits a complex intelligent computing task, such as image recognition and analysis, on a domestically developed desktop computer using a domestically developed desktop client, their raw data is strictly protected in a secure isolation module and intelligently scheduled to the cloud, edge, or other idle desktop nodes for collaborative computing by the privacy computing system. Throughout the computing process, even if a network node fails, the communication link can automatically switch and recover seamlessly. Furthermore, the entire privacy computing system undergoes rigorous intelligent testing and optimization on a verification platform before deployment, ensuring reliability even after large-scale deployment.
[0054] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Those skilled in the art will clearly understand that the technologies in the embodiments of this invention can be implemented using software and necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments of this invention.
[0055] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.
Claims
1. A distributed credit privacy computing system, characterized in that, Comprise: A security isolation module, which adopts a layered architecture design to divide the privacy computing system into a data layer, a computing layer and an application layer, the data layer is used for encrypted storage of original privacy data, the computing layer is used for executing privacy computing tasks, and the application layer is used for providing a user interaction interface; the layers are isolated and communicated through API interfaces and encryption protocols; A cloud-edge-end collaborative operation framework for carrying out the operation throughout the privacy computing system, and realizing distributed privacy computing task execution through the collaboration of a cloud center management node, an edge node and a trust creation desktop client; A communication management module, which monitors the connection health status in real time through heartbeat monitoring, and realizes fault self-healing of the communication link through a master-backup socket switching strategy; A verification platform based on a white box testing method, which tests and verifies the intelligent algorithm, communication logic and system process to be executed or being executed in a simulation environment.
2. The distributed credit privacy computing system according to claim 1, wherein, The security isolation module further comprises a trust creation database, which provides static data encryption for the data layer, data protection in privacy computing and encryption before transmission for the computing layer, and transmission encryption and identity authentication for the application layer.
3. The distributed credit privacy computing system of claim 1, wherein, The cloud center management node performs global task scheduling and resource allocation through a task orchestrator, and aggregates the global model of federated learning in the privacy computing process through a model aggregation server; The edge node connects the cloud center management node and the trust creation desktop client through an edge computing gateway, and preliminarily aggregates the data or model of multiple desktop clients; The trust creation desktop client executes the computing task in the desktop environment through a local task executor, and maintains the state synchronization with the cloud center management node and the edge node through a state synchronizer.
4. The distributed credit privacy computing system of claim 1, wherein, The communication management module comprises a connection management service component, a fault detection and switching engine, and a trust creation socket API; The fault detection and switching engine periodically checks the health status of the master socket and the backup socket through a heartbeat monitoring mechanism, and makes a master-backup socket switching decision when the health status is abnormal; The trust creation socket API executes the master-backup socket switching decision; The connection management service component integrates a pairing table of the master socket and the backup socket and a scheduler, and can record the master-backup relationship and the health status of all connections in real time; the scheduler schedules the master socket and the backup socket to different physical nodes or different memory spaces when a pair of master socket and backup socket needs to be created.
5. The distributed credit privacy computing system according to claim 4, wherein, When the health status is abnormal, anti-ping-pong switching logic is introduced, when the heartbeat detection of the master socket or the backup socket is initially abnormal, the link quality of the master socket or the backup socket is continuously monitored within a set time, and a switching decision or a switching decision is issued.
6. The distributed credit privacy computing system of claim 1, wherein, The verification platform based on a white box testing method, which tests and verifies the intelligent algorithm, communication logic and system process to be executed or being executed in a simulation environment, specifically: In the simulation environment, the input data or training data of the privacy computing model with boundary condition errors or numerical instability problems, the communication network sequence with standby switching failure or causing ping-pong switching, and the node failure timing causing the privacy computing system to appear deadlock or state inconsistency are searched independently to optimize the intelligent algorithm, find the vulnerabilities of communication logic and system flow.
7. The distributed credit privacy computing system according to claim 6, wherein, Maximizing the fitness function The fireworks algorithm is used to continuously generate data that triggers boundary condition errors or data instability problems in the privacy computing model, including privacy computing model training data batches and privacy computing model hyperparameters. ； wherein, , and are the number of times of occurrence of non-number NaN, the number of times of occurrence of gradient explosion and the mutation amplitude value of the loss function value of the privacy calculation model respectively, , and are , and corresponding weights.
8. The distributed credit privacy computing system of claim 6, wherein, Maximizing a fitness function The fireworks algorithm is used to search for communication network sequences that exist standby switching failure or trigger ping-pong switching, including delay time, packet loss rate and fault duration. ； in, , and These represent the number of communication network sequence handover failures, the number of ping-pong handovers per unit time, and the average fault recovery time, respectively. , and They are respectively , and The corresponding weights.
9. The distributed credit privacy computing system of claim 6, wherein, Maximizing a fitness function searching for node fault timing, including fault node, fault type and fault time point, using fireworks algorithm when privacy computing system appears deadlock or state inconsistency ； in, , and These represent the number of times the privacy-preserving computation system experienced a complete deadlock, the number of nodes with inconsistent states, and the percentage decrease in the integrity rate of privacy-preserving computation tasks. , and They are respectively , and The corresponding weights.