An adaptive privacy computing platform and an implementation method thereof

Abstract

This invention relates to the field of privacy computing technology, and discloses an adaptive privacy computing platform and its implementation method. The platform includes several privacy computing nodes and an optional management node. The management node is connected to an optional privacy computing management information persistence component. Each privacy computing node is connected to a computing data persistence component and an optional management data persistence component. By configuring the deployment of the management node, the privacy computing management information persistence component, and the management data persistence component, as well as the subscription and publication of message queues by the management node and the privacy computing nodes, the platform can switch between star and mesh topologies. This helps privacy computing developers reduce development costs while significantly enhancing the platform's application scope and applicable scenarios.

Description

Technical Field

[0001] This invention relates to the field of privacy computing technology, specifically to an adaptive privacy computing platform and its implementation method. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Large internet companies and startups are actively developing privacy computing, launching numerous privacy computing platform products. Due to the inherently distributed nature of privacy computing, current privacy computing platform architectures mainly fall into two categories: star topologies with independent management nodes, and mesh topologies where management functions are attached to computing nodes. Each architecture has its advantages and disadvantages, suitable for different customer needs and hardware / software environments. An architecture that can freely switch between these two modes based on user needs, adapting to different technologies and business environments, and leveraging the advantages of both modes would be of significant importance to the research, development, and promotion of privacy computing platforms.

[0004] The star topology of a privacy computing platform, such as Figure 1 As shown, this architecture has an independent management node and compute nodes belonging to each participant in privacy computing. Management data is stored in the persistence layer of the management node. Compute nodes are only responsible for executing computation tasks and managing local sensitive data and models. Compute nodes receive computation instructions from the management node and retrieve management data from it. From a business perspective, this architecture is suitable for scenarios where there is a leading or managing party in privacy computing. For example, in government data sharing scenarios, the government is generally the managing party; in data trading platform scenarios, the exchange is generally the leading party; and privacy computing platforms provided by cloud service providers also generally adopt this architecture. In such cases, the leading and managing parties usually want an independent management platform to provide stronger control and oversight capabilities. From a technical perspective, the advantages of a star topology are its simple structure, low development cost, ease of management, easy data consistency assurance, and relatively easy fault diagnosis. The disadvantage is that the management node can easily become a system bottleneck.

[0005] The mesh architecture of privacy computing platforms, such as Figure 2As shown, this architecture lacks a dedicated management node; instead, it consists of computing nodes belonging to each participant in the privacy-preserving computation. Management functions are attached to these computing nodes, and management data is stored in the persistence layer of these nodes. Besides sending computation commands, different computing nodes also need to synchronize management data. The primary driver for this architecture is the business scenario. In many privacy-preserving computation scenarios, all parties are fully equal business entities, such as data sharing among multiple banks. Deploying a management node in this context would raise concerns about data security if it were deployed by any participant or third party. Furthermore, in some cases, it is difficult to find a location on the network that is connected to the computing nodes of all participants to place the management node. From a technical perspective, the advantage of a mesh architecture is the absence of a single point of failure that could become a system bottleneck. However, its disadvantages include high development costs, complex management logic, and difficulty in ensuring data consistency across nodes. Summary of the Invention

[0006] To address the aforementioned problems, this invention provides an adaptive privacy computing platform and its implementation method. This platform can switch between star and mesh architectures based on different business scenarios and hardware / software environment conditions. This helps privacy computing developers significantly enhance the platform's application scope and applicable scenarios while reducing development costs.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A first aspect of the present invention provides an adaptive privacy computing platform comprising:

[0009] It includes several privacy computing nodes and an optional management node; the management node is connected to an optional privacy computing management information persistence component; the privacy computing node is connected to a computing data persistence component and an optional management data persistence component;

[0010] By monitoring the deployment of management nodes, privacy computing management information persistence components, and management data persistence components, as well as the configuration of management nodes and privacy computing nodes for subscribing to and publishing message queues, the system can switch between star and mesh topologies.

[0011] Furthermore, the conditions for switching to star mode are:

[0012] Deploy management nodes and privacy computing management information persistence components, but do not deploy management data persistence components;

[0013] Compute nodes subscribe to their own management message queues and instruction message queues. Management nodes control and schedule compute nodes by publishing messages to the management message queues and instruction message queues of the compute nodes.

[0014] Furthermore, the conditions for switching to mesh mode are:

[0015] Instead of deploying management nodes and privacy computing management information persistence components, a management data persistence component is deployed.

[0016] Compute nodes can subscribe to and publish management message queues and instruction message queues on an equal footing.

[0017] Furthermore, before joining or leaving the privacy computing network, the management node and the privacy computing node send a networking message to the networking queue.

[0018] Furthermore, the networking message includes node ID, node type, and node networking action.

[0019] Furthermore, in the star topology, the management node creates a privacy computing task. After the participating nodes of the privacy computing task confirm the privacy computing task, the task management data of the privacy computing task is stored on the management node.

[0020] Furthermore, in the star topology, after a privacy computing node starts a privacy computing task, the management node publishes a task start message to the instruction message queue of the privacy computing task initiating node. After receiving the task start message, the privacy computing node starts the execution of the privacy computing task, and after the task is completed, it publishes a task status update message to the management message queue. After receiving the task status update message, the management node updates the local task status.

[0021] Furthermore, in the mesh architecture, privacy computing node A creates a privacy computing task. Privacy computing node A sends a confirmation message for the privacy computing task to the management message queue of participating privacy computing node B. After receiving the message, privacy computing node B confirms its participation in the execution of the privacy computing task and sends a task confirmation message to the management message queue of privacy computing node A. The management data of the privacy computing task is simultaneously stored in privacy computing node A and privacy computing node B.

[0022] Furthermore, in the mesh architecture, after privacy computing node A starts a privacy computing task, it publishes the task start message to the instruction message queue of privacy computing node B. Upon receiving the task start message, privacy computing node B also starts the privacy computing task. After the privacy computing task is completed by privacy computing nodes A and B, privacy computing node A publishes a task status update message to the management message queue of privacy computing node B, and privacy computing node B also publishes a task status update message to the management message queue of privacy computing node A. After receiving the task status update message, privacy computing nodes A and B update their local task status respectively.

[0023] A second aspect of the present invention provides a method for implementing an adaptive privacy computing platform as described in the first aspect, comprising the following steps:

[0024] The system obtains information on the deployment of the management node, the privacy computing management information persistence component, and the management data persistence component, as well as the configuration of the management node and privacy computing node for subscribing to and publishing message queues, enabling switching between star and mesh topologies.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] This invention provides an adaptive privacy computing platform that can switch between star or mesh architectures according to different business scenarios and hardware and software environment conditions. This helps privacy computing developers reduce development costs while significantly enhancing the platform's application scope and applicable scenarios. Attached Figure Description

[0027] The accompanying drawings, which constitute a part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0028] Figure 1 Star architecture diagram for existing privacy computing platforms;

[0029] Figure 2 A mesh architecture diagram of an existing privacy computing platform;

[0030] Figure 3 This is an architecture diagram of an adaptive privacy computing platform according to Embodiment 1 of the present invention;

[0031] Figure 4 This is a message queue configuration diagram of a privacy computing platform in star topology according to Embodiment 1 of the present invention;

[0032] Figure 5 This is a message queue configuration diagram of a privacy computing platform in mesh mode according to Embodiment 1 of the present invention;

[0033] Figure 6 This is a flowchart illustrating the process of node joining and leaving using a network queue, as described in Embodiment 1 of the present invention.

[0034] Figure 7 This is a flowchart of the privacy computing task processing in the star schema of Embodiment 1 of the present invention;

[0035] Figure 8 This is a flowchart illustrating the privacy computing task processing in the mesh pattern of Embodiment 1 of the present invention. Detailed Implementation

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0037] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0038] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0039] Unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0040] Example 1

[0041] The purpose of this first embodiment is to provide an adaptive privacy computing platform.

[0042] This embodiment provides an adaptive privacy computing platform that addresses the advantages and disadvantages of the two privacy computing architectures mentioned above. It proposes an adaptive architecture that allows products based on this architecture to be deployed in a traditional star topology or mesh architecture as needed during the deployment phase.

[0043] This embodiment provides an adaptive privacy computing platform that can switch between a star topology and a mesh topology based on different business scenarios and hardware / software environment conditions. Users only need to configure the architecture during product deployment to enable the privacy computing platform to run in either a star or mesh topology. This adaptive privacy computing platform architecture helps privacy computing developers reduce development costs while significantly enhancing the platform's application scope and applicable scenarios.

[0044] This embodiment provides an adaptive privacy computing platform, such as Figure 3As shown, a distributed message queue component acts as the intermediary for interaction between nodes, combining the characteristics of star and mesh architectures. In this embodiment, the components represented by dashed lines in the adaptive privacy computing platform architecture are optional. For example, the management node and its connected privacy computing management information persistence layer are optional components, as is the management data persistence layer connected to the privacy computing node. When a management node is deployed, this architecture transforms into a star topology. The management node manages the privacy computing nodes (hereinafter referred to as computing nodes) by sending management messages and computation instruction messages to the distributed message queue. The computing nodes subscribe to and process messages in their respective message queues. The computing nodes themselves store local computation data but do not store management data. In the absence of a management node and with the configuration of a component for persisting local management information (management data persistence) on the compute nodes, the architecture transforms into a mesh architecture. In this case, participating nodes in the privacy computing task implement management functions by sending management messages and computing instruction messages to a distributed message queue. Each participating node subscribes to and processes messages in its own message queue to achieve mutual computing collaboration and synchronization of management information data. The compute nodes themselves store both local computing data and management data.

[0045] The privacy computing management information persistence layer is used to store relevant management information in the privacy computing scenario; it is essentially a database for storing management information. For example, when a user creates a privacy computing task, the persistence layer is responsible for storing relevant information about that task (participants, participating data, scheduling information, result recipients, etc.). The management data persistence component stores management control data and provides read / write functionality for it. The computation data persistence component stores computation data and provides read / write functionality for it. For example, management control data includes privacy computing task information, node information, user information, data authorization information, etc. Computation data refers to the data used and generated during the specific execution of the privacy computing task, such as the floating-point numbers provided by each party and the result of the addition calculation when performing a multi-party secure addition task.

[0046] 1. Message types between nodes.

[0047] In the architecture of this adaptive privacy computing platform, a distributed message queue is located at the core. Inter-node interaction and data synchronization are all achieved through the message queue. Messages sent and received between management nodes and computing nodes, and between different computing nodes, can be divided into the following two types:

[0048] (1) Privacy Computing Management Messages: These messages mainly involve reading and writing metadata related to privacy computing, and are unrelated to the execution of specific privacy computing tasks. They can be further divided into the following categories:

[0049] Privacy computing task management messages, such as messages related to task creation, confirmation, modification, freezing, activation, and deletion;

[0050] Privacy-preserving computing node management messages, such as node registration, activation, heartbeat, modification, and deletion;

[0051] Privacy Computing Consortium (Group) management messages, such as the creation, confirmation, and node joining / leaving of the consortium;

[0052] Data management messages, such as data template creation, data upload, authorization, etc.;

[0053] Model management messages, such as model creation, publishing, and canceling publishing.

[0054] (2) Calculation instruction messages:

[0055] Instructions for privacy-preserving computing tasks, such as task start / execution and termination messages;

[0056] Data manipulation commands, such as uploading data and retrieving results;

[0057] Model manipulation commands, such as model import, export, and calling model services.

[0058] 2. Message publishing and subscription in star schema.

[0059] The adaptive architecture proposed in this invention determines whether it operates in star or mesh topology by changing the following two configurations:

[0060] Whether to deploy management nodes and their associated components, and whether to deploy management information persistence components for compute nodes;

[0061] Configuration of message queues subscribed to by management nodes and compute nodes, and queues for publishing messages.

[0062] The adaptive architecture proposed in this invention switches to star mode operation through the following two configurations:

[0063] Deploy management nodes and their associated components, but do not deploy management information persistence components for compute nodes;

[0064] Compute nodes subscribe to their own management message queue and instruction message queue. Management nodes control and schedule compute nodes by publishing messages to the compute nodes' management message queue and instruction message queue.

[0065] In star topology, all management and control information is stored on the management node, while the compute nodes do not store this information.

[0066] In star topology, the message queue configurations related to management nodes and compute nodes, such as... Figure 4 As shown in the diagram, this configuration allocates two dedicated message queues to each compute node: a management message queue and an instruction message queue. The management message queue is further divided into queues for tasks, nodes, data, and models, based on the managed object. Similarly, the instruction message queue is divided into queues for tasks, nodes, data, and models, based on the object of the instruction operation. Compute nodes subscribe to their own management and instruction message queues. The management node controls and schedules nodes by publishing messages to the specific node's management and instruction message queues. For example, the management node publishes a node certificate update message to compute node A's management message queue; upon receiving this message, node A updates its local certificate. Similarly, the management node publishes a message to compute node A's instruction message queue to start a privacy computing task; upon receiving this message, node A begins running the privacy computing task.

[0067] The management node also has its own set of message queues, including queues for tasks, nodes, data, and models. All compute nodes publish management messages to these queues, and the management node subscribes to these message queues, receives messages from them, and executes corresponding management operations. For example, after a compute node successfully completes a privacy computing task, it sends a message to the management node's task message queue to update the task status. Upon receiving this message, the management node sets the task's execution status to successful.

[0068] 3. Message publishing and subscription in mesh mode.

[0069] The adaptive architecture proposed in this invention switches to mesh mode through the following two configurations:

[0070] The management node and its associated components are not deployed, but the management information persistence component of the computing node is deployed.

[0071] Compute nodes can subscribe to and publish configurations for peer-to-peer message queues.

[0072] In mesh mode, the message queue configuration related to compute nodes, such as Figure 5As shown in the diagram, this configuration allocates two dedicated message queues to each compute node: a management message queue and an instruction message queue. Compute nodes subscribe to their own management and instruction message queues and simultaneously publish management and instruction messages to other cooperating compute nodes. For example, compute node A publishes a message to compute node B's management message queue to create a privacy computing task. After receiving this message, node B updates its local task management information and decides whether to participate in the task. If it participates, it publishes a task confirmation message to node A's management message queue. Upon receiving this message, node A updates its local task management information (in this configuration, a persistent management information component is deployed locally on the node).

[0073] 4. Nodes joining and leaving the privacy computing platform.

[0074] In the adaptive privacy computing platform architecture proposed in this invention, a network queue is designed to support the automatic joining and leaving of computing nodes, such as... Figure 6 As shown, this network queue is subscribed to by all compute nodes and management nodes (if any). Before joining or leaving the privacy computing network, a node sends a network message to the network queue, which includes the node ID, node type, and node networking action (joining / leaving). After receiving the message, other nodes that have subscribed to the network message update their message subscription list according to the message content, adding or canceling the corresponding subscription rules to achieve automatic networking of nodes.

[0075] 5. Message processing flow in a star topology.

[0076] Based on the aforementioned adaptive architecture and distributed message publishing and subscription rules, a message processing flow between the components of the privacy computing platform was designed. For the message processing flow in a star topology, the following example, using a privacy computing task processing scenario, illustrates the specific message flow between nodes and the internal processing logic of each node. This flow is as follows: Figure 7 As shown.

[0077] The process begins with an administrator on a compute node creating a privacy-preserving computation task on the management node. Administrators on other participating nodes then confirm the task, and all relevant task management data is stored on the management node. Next, the user initiates the task. The management node then publishes a task start message to the corresponding instruction message queue of the task-initiating node. Since the compute node has subscribed to this queue, it starts the task upon receiving the message (the instruction message transmission process between different compute nodes during execution depends on the specific privacy-preserving computation protocol and is omitted here). After the task finishes, the node publishes a task status update message to the management node's message queue. Since the management node has subscribed to this queue, it updates its local task status upon receiving the message. The user can then query the updated task status through the management node.

[0078] 6. Message processing flow in a mesh architecture.

[0079] For the message processing flow of an adaptive architecture in mesh mode, the following example, using a privacy computing task processing scenario, illustrates the specific message flow between nodes and the processing logic within each node. The flow is as follows: Figure 8 As shown.

[0080] The process begins with the administrator of compute node A creating a privacy-preserving computation task on node A. Since node B is a participating node in this task, node A sends a task confirmation message to the management message queue of participating node B. Upon receiving this message, participating node B confirms its participation in the execution of the task and then sends a task confirmation message to the management message queue of node A. At this point, all management data related to the task is stored on both nodes A and B. Next, the administrator of node A starts the execution of the task. Node A then publishes the task start message to the instruction message queue of node B. Since compute node B has subscribed to this queue, upon receiving the message, node B also starts the execution of the task (the process of instruction message transmission between different compute nodes during execution is related to the specific privacy-preserving computation protocol, and this part is omitted). After the task finishes running on nodes A and B, node A publishes a task status update message to the management message queue of node B, and node B also publishes a task status update message to the management message queue of node A. After receiving the messages, nodes A and B update their local task status, and the task status remains synchronized between the two nodes. Users can query the latest status of the task through nodes A or B.

[0081] The adaptive privacy computing platform proposed in this invention can flexibly switch between star and mesh topologies, helping privacy computing developers use a single architecture to address both dominant, heavily regulated privacy computing scenarios and scenarios emphasizing equal status and shared responsibility among all participants.

[0082] The adaptive privacy computing platform proposed in this invention reduces development costs while significantly improving the adaptability of the privacy computing platform, enabling privacy computing platform developers to adapt a single architecture to different business needs and software and hardware environments.

[0083] Example 2

[0084] The purpose of this second embodiment is to provide an implementation method for an adaptive privacy computing platform as described in the first embodiment, including the following steps:

[0085] The system obtains information on the deployment of the management node, the privacy computing management information persistence component, and the management data persistence component, as well as the configuration of the management node and privacy computing node for subscribing to and publishing message queues, enabling switching between star and mesh topologies.

[0086] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0087] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. An adaptive privacy computing platform, characterized by: It includes several privacy computing nodes and an optional management node; the management node is connected to an optional privacy computing management information persistence component; the privacy computing node is connected to a computing data persistence component and an optional management data persistence component; By monitoring the deployment of management nodes, privacy computing management information persistence components, and management data persistence components, as well as the configuration of message queue subscriptions and publications on management nodes and privacy computing nodes, the system enables switching between star and mesh topologies. The conditions for switching to star topology are: Deploy management nodes and privacy computing management information persistence components, but do not deploy management data persistence components; Compute nodes subscribe to their own management message queues and instruction message queues. Management nodes control and schedule compute nodes by publishing messages to the management message queues and instruction message queues of the compute nodes. The conditions for switching to mesh mode are: Do not deploy management nodes and privacy computing management information persistence components, but deploy management data persistence components; Compute nodes can subscribe to and publish management message queues and instruction message queues on an equal footing.

2. The adaptive privacy computing platform as described in claim 1, characterized in that: Before joining or leaving the privacy computing network, the management node and privacy computing node send a networking message to the networking queue.

3. The adaptive privacy computing platform as described in claim 2, characterized in that: The networking message includes node ID, node type, and node networking action.

4. The adaptive privacy computing platform as described in claim 1, characterized in that: In a star topology, the management node creates a privacy computing task. After the participating nodes of the privacy computing task confirm the privacy computing task, the task management data of the privacy computing task is stored on the management node.

5. An adaptive privacy computing platform as described in claim 1, characterized in that: In a star topology, after a privacy computing node starts a privacy computing task, the management node publishes a task start message to the instruction message queue of the privacy computing task initiating node. After receiving the task start message, the privacy computing node starts the execution of the privacy computing task, and after the task finishes running, publishes a task status update message to the management message queue. After receiving the task status update message, the management node updates the local task status.

6. The adaptive privacy computing platform as described in claim 1, characterized in that: In a mesh architecture, privacy computing node A creates a privacy computing task. Privacy computing node A sends a confirmation message for the privacy computing task to the management message queue of participating privacy computing node B. After receiving the message, privacy computing node B confirms its participation in the execution of the privacy computing task and sends a task confirmation message to the management message queue of privacy computing node A. The management data of the privacy computing task is stored in both privacy computing node A and privacy computing node B.

7. An adaptive privacy computing platform as described in claim 1, characterized in that: In a mesh architecture, after privacy computing node A starts a privacy computing task, it publishes the task start message to the instruction message queue of privacy computing node B. Upon receiving the task start message, privacy computing node B also starts the privacy computing task. After the privacy computing task is completed by privacy computing nodes A and B, privacy computing node A publishes a task status update message to the management message queue of privacy computing node B, and privacy computing node B also publishes a task status update message to the management message queue of privacy computing node A. After receiving the task status update messages, privacy computing nodes A and B update their local task status respectively.

8. A method for implementing an adaptive privacy computing platform as described in any one of claims 1-7, characterized in that: Includes the following steps: The system obtains information on the deployment of the management node, the privacy computing management information persistence component, and the management data persistence component, as well as the configuration of the management node and privacy computing node for subscribing to and publishing message queues, enabling switching between star and mesh topologies.

Images