WO2022267787A1

WO2022267787A1 - Method, apparatus, and system for determining computation resource in privacy computation

Info

Publication number: WO2022267787A1
Application number: PCT/CN2022/094323
Authority: WO
Inventors: 应鹏飞; 李漓春; 王一凡
Original assignee: 支付宝(杭州)信息技术有限公司
Priority date: 2021-06-25
Filing date: 2022-05-23
Publication date: 2022-12-29
Also published as: CN115525919A

Abstract

Embodiments of the present description provide a method for determining a computation resource in privacy computation. The method is executed by a management node. The method comprises: receiving a query request for a computation resource from a computation resource usage demander, wherein the computation resource is a public resource supporting a privacy algorithm, and the computation resource is pre-registered in a management node; determining a resource list on the basis of the query request and the resource information of a plurality of locally registered computation resources; and transmitting the resource list to the computation resource usage demander, so that the computation resource usage demander determines, from the resource list, a target resource used for executing a target privacy computation task.

Description

Method, device and system for determining computing resources in privacy computing

technical field

One or more embodiments of this specification relate to the field of computer technology, and in particular to methods, devices and systems for determining computing resources in privacy computing.

Background technique

In order to fully tap the value of data, multi-party joint data calculation has become a major research hotspot. For example, in order to better classify and analyze merchants, the electronic payment platform will classify merchants based on the merchant transaction data it owns, the e-merchant platform based on the merchant sales data it owns, and the banking institution based on the merchant loan data it owns. analyze.

In multi-party data joint computing, in order to protect the data privacy of all parties, it is necessary to use Privacy-Preserving Computation technology. Privacy-preserving computing has a variety of computing routes, such as Secure Multi-Party Computation (MPC), Trusted Execution Environment (Trusted Execution Environment, TEE), Federated Learning (Federated Learning, FL), etc. In practical applications, privacy computing routes can be selected for specific privacy computing tasks.

Contents of the invention

One or more embodiments of this specification describe the method for determining computing resources in privacy computing. By introducing management nodes to establish and manage computing resource pools, the holders and demanders of computing resources can release resources conveniently and quickly. and discover.

According to the first aspect, there is provided a method for determining computing resources in privacy computing, the method is executed by a management node, and the method includes: receiving a query request for computing resources from a computing resource usage demand side, wherein the computing resources support privacy The public resources of the algorithm, the computing resources are pre-registered in the management node; based on the query request and the resource information of multiple computing resources registered locally, determine the resource list; send the resources to the computing resource usage demand side list, so that the computing resource usage demand side determines target resources from the resource list for executing target privacy computing tasks.

In one embodiment, the query request includes the target privacy algorithm adopted for the target privacy computing task, and the resource information includes the privacy algorithm supported by the computing resource; wherein, based on the query request and locally registered multiple resource information of computing resources, and determining a resource list includes: determining a number of computing resources supporting the target privacy algorithm to form the resource list.

In a specific embodiment, the method further includes: including the current usage state information of the several computing resources in the resource list.

In one embodiment, the resource information includes privacy algorithms supported by computing resources; wherein, based on the query request and the resource information of a plurality of locally registered computing resources, determining a resource list includes: determining the plurality of computing resources Among the resources, several computing resources that are currently in an idle state; the resource list is determined according to the several computing resources and the privacy algorithm supported by each computing resource.

In one embodiment, after sending the resource list to the computing resource usage demander, the method further includes: receiving a first notification from the computing resource usage demander, which indicates that the target resource is used by the Computing resources used by the demand side; according to the first notification, updating the usage status of the target resource in the resource information.

In a specific embodiment, after receiving the first notification from the computing resource usage demander, the method further includes: receiving a second notification from the computing resource usage demander, which indicates the computing resource usage demand The party stops using the target resource; according to the second notification, update the usage status of the target resource in the resource information; The cost incurred by the resource.

In one embodiment, the resource list includes IP addresses of each computing resource therein.

In one embodiment, before receiving a query request for computing resources from a computing resource usage demander, the method further includes: receiving a registration request from a computing resource holder, including registration information for computing resources, the registration The information includes a privacy algorithm supported by a corresponding computing resource; and if the registration request is approved, the registration information is included in the resource information.

In a specific embodiment, adding the registration information to the resource information includes: allocating a resource ID to the computing resource corresponding to the registration request; and associating and storing the resource ID and the registration information.

In a more specific implementation, the registration information further includes at least one of the following: IP address, domain name, and public key of the corresponding computing resource.

According to the second aspect, there is provided a method for determining computing resources in privacy computing, the method is executed by a demander for computing resources, and the method includes: sending a query request for computing resources to a management node, where the computing resources support privacy public resources of the algorithm, the computing resources are pre-registered at the management node; receiving a resource list from the management node, the resource list is determined based on the query request and the resource information of a plurality of locally registered computing resources; based on the The above resource list is used to determine the target resource used to perform the target privacy computing task.

In an embodiment, the method further includes: determining a target privacy algorithm adopted for the target privacy computing task, and including the target privacy algorithm in the query request.

In a specific embodiment, determining a target resource for executing a target privacy computing task based on the resource list includes: randomly selecting a computing resource from the resource list as the target resource.

In another specific embodiment, the resource list includes current usage status information of each computing resource; wherein, based on the resource list, determining the target resource for executing the target privacy computing task includes: from the resource A computing resource whose current use state is idle is selected from the list as the target resource.

In an embodiment, the resource list includes privacy algorithms supported by each computing resource; wherein, based on the resource list, determining the target resource for executing the target privacy computing task includes: based on the target privacy computing For the target privacy algorithm adopted by the task, a computing resource supporting the target privacy algorithm is selected from the resource list as the target resource.

In one embodiment, after determining the target resource for executing the target privacy computing task, the method further includes: sending the IP address of the target resource to other participants; utilizing the target resource, communicating with the other Participants jointly perform the target privacy computing task.

According to a third aspect, a method for determining computing resources in privacy computing is provided, including: sending a query request for computing resources to a management node from a computing resource usage demand direction, wherein the computing resources are public resources that support privacy algorithms, and the computing resources are pre-determined Registered at the management node; the management node determines a resource list based on the query request and the resource information of multiple computing resources registered locally; the management node sends the resource list to the computing resource usage demand side ; The computing resource usage demander determines the target resource for executing the target privacy computing task based on the resource list.

In one embodiment, the method further includes: the computing resource usage demand side determines the target privacy algorithm adopted for the target privacy computing task, and includes the target privacy algorithm in the query request; wherein, the management The node determines a resource list based on the query request and resource information of multiple computing resources stored therein, including: determining several computing resources supporting the target privacy algorithm to form the resource list.

In a specific embodiment, the method further includes: the management node includes the current usage status information of the several computing resources in the resource list; list, determining a target resource for executing a target privacy computing task, comprising: selecting a computing resource whose current use state is idle from the resource list as the target resource.

In yet another specific embodiment, the above-mentioned determination of several computing resources that support the target privacy algorithm and forming the resource list includes: the management node, based on the current usage status information of the several computing resources, from the several computing resources Selecting computing resources whose current use status is idle from the resources to form the resource list; wherein, the computing resource use demand side determines the target resource for executing the target privacy computing task based on the resource list, including: the The computing resource use demand side determines the computing resource in the resource list as the target resource.

In another specific embodiment, the management node determines the resource list based on the query request and the resource information of multiple computing resources stored in it, including: determining several of the multiple computing resources that are currently idle Computing resources to form the resource list; wherein, based on the resource list, determining the target resource for executing the target privacy computing task includes: based on the target privacy algorithm adopted for the target privacy computing task, from the resource list Select a computing resource that supports the target privacy algorithm as the target resource.

According to the fourth aspect, there is provided a device for determining computing resources in privacy computing, the device is integrated in a management node, and the device includes: a query request receiving unit configured to receive a query request for computing resources from a computing resource usage demander , wherein the computing resource is a public resource that supports a privacy algorithm, and the computing resource is pre-registered at the management node; the resource list determination unit is configured to determine the resource based on the query request and resource information of a plurality of locally registered computing resources list; a resource list sending unit configured to send the resource list to the computing resource usage demander, so that the computing resource usage demander can determine target resources from the resource list for executing target privacy computing tasks.

According to the fifth aspect, there is provided a device for determining computing resources in privacy computing, the device is integrated with a computing resource usage demander, and the device includes: a query request sending unit configured to send a query request for computing resources to a management node , wherein the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered at the management node; the resource list receiving unit is configured to receive a resource list from the management node, the resource list is based on the query request and the local The resource information of a plurality of registered computing resources is determined; the target resource determining unit is configured to determine a target resource for executing a target privacy computing task based on the resource list.

According to the sixth aspect, a system for determining computing resources in privacy computing is provided, including: a demander for computing resources, configured to send a query request for computing resources to a management node, where the computing resources are public resources that support privacy algorithms, so The computing resources are pre-registered on the management node; the management node is configured to determine a resource list based on the query request and the resource information of multiple computing resources registered locally, and send the resource list to the computing resource usage demand party The resource list; the computing resource usage demander is further configured to determine a target resource for executing a target privacy computing task based on the resource list.

According to a seventh aspect, there is provided a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed in a computer, the computer is made to execute the method provided in any one of the first to third aspects .

According to an eighth aspect, there is provided a computing device, including a memory and a processor, wherein executable code is stored in the memory, and when the processor executes the executable code, any one of the first aspect to the third aspect is implemented method provided.

According to the method and device provided by the embodiments of this specification, the management node supports the computing resource holder to register the computing resources it holds, thereby establishing a resource pool, so that the demander with computing resource usage requirements can submit to the management node Send a resource query request, and determine the target resource used to perform the target privacy computing task based on the resource list returned by the management node, and then the computing resource usage demander can cooperate with other participants to jointly use the target resource to complete the target privacy computing task execution. In this way, by introducing management nodes to establish and manage computing resource pools, the holders and demanders of computing resources can release and discover resources conveniently and quickly.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For Those of ordinary skill in the art can also obtain other drawings based on these drawings without making creative efforts.

Fig. 1 shows a schematic diagram of an interaction scenario for computing resources according to an embodiment;

Fig. 2 shows a schematic diagram of multi-party interaction in determining computing resources in privacy computing according to an embodiment;

Fig. 3 shows a schematic diagram of multi-party collaborative privacy calculation according to an embodiment;

Fig. 4 shows a schematic diagram of multi-party collaborative privacy calculation according to another embodiment;

Fig. 5 shows a schematic diagram of multi-party interaction in determining computing resources in privacy computing according to another embodiment;

FIG. 6 shows a schematic diagram of an architecture for performing privacy data statistics tasks according to an embodiment;

Fig. 7 shows a schematic structural diagram of an apparatus for determining computing resources in privacy computing according to an embodiment;

Fig. 8 shows a schematic structural diagram of an apparatus for determining computing resources in privacy computing according to another embodiment.

detailed description

The solutions provided in this specification will be described below in conjunction with the accompanying drawings.

In order to complete a privacy computing task, it may be necessary to use computing resources provided by a third party, such as MPC random number server or TEE server. For the selection of third parties and computing resources, inquiries and negotiations are often conducted in an offline point-to-point manner. This specification provides another solution to determine computing resources in private computing.

The solution proposed by the inventor introduces management nodes to collect computing resources released by third parties, and establishes computing resource pools, so that demanders who have computing resource usage requirements can send resource query requests to management nodes, and based on the management nodes according to Query the resource list returned by the request to determine the target resource used to perform the target privacy computing task, and then, the demander can inform each participant of the privacy computing of the addressing information of the target resource, so that each participant can jointly use the The target resource completes the execution of the target privacy computing task.

To facilitate intuitive understanding, FIG. 1 shows a schematic diagram of an interaction scenario for computing resources according to an embodiment. As shown in Figure 1, bank A sends a query request to the management node, including its request for computing resources to support the TEE algorithm; the management node returns a list of computing resources to bank A based on the resource information of each resource in its computing resource pool, including 3 The usage status of each TEE server; then, bank A selects No. 3 TEE server in an idle state, and sends the Internet Protocol (Internet Protocol, IP) address of No. 3 TEE server to bank B, and then uses it with bank B No. 3 TEE server completes the target privacy calculation task (for example, jointly establishes a merchant classification model).

Below, in conjunction with specific embodiments, the implementation steps for realizing the above inventive concept will be described. Fig. 2 shows a schematic diagram of multi-party interaction in determining computing resources in privacy computing according to an embodiment, and the multi-party at least includes a computing resource usage demander and a management node. Exemplarily, the computing resource usage demander may be bank A shown in FIG. 1 . It should be understood that each of the multiple parties can be implemented as a device, platform, server or device cluster with computing and processing capabilities.

As shown in FIG. 2 , the interaction process includes the following steps: Step S210 , the computing resource usage demander sends a query request for the computing resource to the management node, and the computing resource is registered in the management node in advance.

It should be noted that computing resources are public resources that support privacy algorithms, where relevant computing logic for privacy algorithms is deployed, and/or are used to provide relevant algorithm parameters, which can be shared by multiple privacy computing tasks. Correspondingly, the computing resource can be carried by the public resource platform, and the public resource platform assumes the role of the algorithm provider in the process of publishing and providing the privacy algorithm; The resource platform also assumes the role of computing party. Because computing resources are public resources, it means that the public resource platform only needs to provide computing resources, and it does not directly or indirectly provide sample data related to privacy computing, and the public resource platform can use the computing resources it provides , to flexibly establish cooperative relationships with different, non-specific privacy computing participants in multiple privacy calculations, and jointly perform privacy calculations.

In one implementation, the data provider in the privacy calculation relies on the calculation party with the privacy calculation engine to perform privacy calculation. Specifically, the data provider sends the private sample data held by itself to the first type of computing resources for privacy computing, for example, after encrypting the sample data and sending it to the computing party for privacy computing; for example, after sharding the sample data Send it to the computing party for privacy calculation. Fig. 3 shows a schematic diagram of multi-party collaborative privacy computing according to an embodiment. As shown in Fig. 3, n data providers perform privacy computing based on computing power provided by t computing nodes. In a specific embodiment, the resources of the t computing nodes may be TEE servers; in another embodiment, the t computing nodes may be servers with MPC computing engines.

In another implementation situation, the data provider itself also participates in the privacy calculation and assumes a part of the role of the calculation party. At this point, the second type of computing resources (or auxiliary computing resources) can be further sought to participate in privacy computing together with it. In some cases, this cooperative privacy computing method can adopt a simplified or better privacy computing protocol. Fig. 4 shows a schematic diagram of multi-party collaborative privacy computing according to another embodiment. As shown in Fig. 4, the two data providers also serve as computing nodes and cooperate with auxiliary computing resources (shown as auxiliary servers in Fig. 4) Perform privacy calculations. In a specific embodiment, the auxiliary computing resource may be an MPC random number provider. In another embodiment, the secondary computing resource may be a Federated Learning Center server.

It should be understood that the carrier of the above computing resources may be any module, device or device cluster with computing and processing capabilities.

On the other hand, in one embodiment, the above query request indicates a request to query computing resources in an available state. In a specific embodiment, being in the available state may include being in a partially idle state or a completely idle state.

In one embodiment, the computing resource usage requester determines the target privacy algorithm adopted for the target privacy computing task, and includes the target privacy algorithm in the query request. It should be noted that there may be multiple types of target privacy computing tasks. In a specific embodiment, the task type may include joint statistical analysis for specified data items. In a more specific embodiment, the specified data items may be user age, user gender, merchant average turnover, merchant category, enterprise scale, and the like. In a more specific embodiment, the joint statistical analysis may be to analyze the numerical distribution of the specified data item within its definition domain, or, between different parties, to compare the value corresponding to the specified data item, or, between different parties Analysis and calculation operations such as summation, intersection, or averaging of specified data items. In an example, the joint statistical analysis for the specified data item may be: counting the total number of female users in different parties. In another example, it may also be: counting the number of identical users in different parties. In yet another example, it may also be: counting age distributions of all users in the multi-party. In yet another example, it may be: comparing the average turnover of merchants of multiple parties.

In another specific embodiment, the task type of the target computing task may include joint training of specified models. In a more specific embodiment, the specified model may be a machine learning model such as a logistic regression model, a decision tree model, or a neural network model. In a more specific embodiment, the specified model may be a classification model or a regression model. In a more specific embodiment, the sample objects targeted by the specified model may be: users, merchants, commodities, events (such as login events or access events), devices (such as user terminals or Internet of Things devices), and so on. In an example, the specified model may be a user classification model or a user scoring model, such as a consumer group segmentation model, a user risk assessment model, a user anomaly identification model, and the like. In another example, the specified model may be a commodity classification model or a commodity scoring model, specifically such as a commodity popularity scoring model.

In yet another specific embodiment, the task type of the target computing task may include joint prediction of a specified model. In a more specific embodiment, the specified model may be a classification model or a regression model, and correspondingly, the joint prediction may be to predict a classification category based on a classification model, or to predict a regression value based on a regression model. In an example, the joint prediction of the above specified models may be: predicting user risk scores based on the user risk assessment model, or identifying abnormal users based on the user anomaly identification model.

The above description describes the target computing task types included in the description information. It should be understood that in addition to the above-mentioned joint statistical analysis for specified data items, joint training and joint prediction of specified models, task types can also include other types, which are not exhaustive.

Generally, one or more privacy algorithms that perform the above target privacy tasks can be determined. For example, assuming that the target privacy task is to determine the number of identical users in multiple service platforms, correspondingly, the determined privacy algorithm may include a Private Set Intersection (PSI) algorithm, a TEE algorithm, and the like. Further, in a specific embodiment, when multiple privacy algorithms are determined, one of them may be randomly selected as the target privacy algorithm. In another specific embodiment, the plurality of privacy algorithms may be prioritized, and the privacy algorithm with the highest priority may be included in the query request as the target privacy algorithm. Furthermore, in an example, the staff sets priorities for all privacy algorithms that may be adopted in advance, so that in this step, the above-mentioned multiple privacy algorithms can be prioritized according to the preset priority information. In another example, the above-mentioned multiple privacy algorithms may be prioritized according to the historical implementation effects of the privacy algorithms.

Based on the above, the computing resource usage demander can determine the target privacy algorithm adopted for the privacy target task, and include it in the above query request. In a specific embodiment, the above query request may also include the required number of computing resources supporting the target privacy algorithm (such as 2 servers, etc.).

The computing resource usage demander sends the generated query request to the management node. Correspondingly, the management node may receive the query request, and in step S220, determine a resource list based on the query request and the resource information of the plurality of computing resources registered locally.

In an embodiment, the resource information may include resource ID (Identity) of the computing resource, addressing information, public key, supported privacy algorithm, and the like. In a specific embodiment, the addressing information may include an IP address. In another specific embodiment, the addressing information may include a domain name.

Regarding the source of resource information, in one embodiment, referring to Figure 5, the above-mentioned multiple parties may also include computing resource holders. Before step S210, the process of multi-party interaction may also include steps S205 and S206. In step S205, the management The node receives the registration request from the computing resource holder, which includes the registration information for the computing resource, and the registration information includes the privacy algorithm supported by the corresponding computing resource; in step S206, when the above registration request is approved, the management node will Registration information falls under the resource information. It should be understood that computing resource holders can become computing resource issuers by registering their own computing resources with management nodes, so that management nodes can establish resource pools based on registered computing resources for query and use by demanders of computing resources . In addition, resource issuers can issue multiple computing resources that support the same or different privacy algorithms.

In a specific embodiment, the above registration information may also include the IP address, domain name or public key of the corresponding resource. In a specific embodiment, the review of the above-mentioned registration request may include conducting qualification review and credit review of the computing resource holder, and reviewing the validity of the computing resources involved in the registration request. In a specific embodiment, the management node adding the registration information to the resource information may include: allocating a resource ID to the computing resource corresponding to the registration request; and storing the resource ID and the registration information in association.

In another embodiment, the above resource information may also include the current usage status of computing resources, for example, in use and idle. In a specific embodiment, usage details may also be included. In a more specific embodiment, the usage details may include: how many merchants or data nodes the computing resource is used by, the CPU, memory, disk, network, etc. occupancy and remaining availability of the computing resource in the last minute. In an example, the management node may obtain the above usage details through active query. In another example, the above usage details may be proactively reported to the management node by the above multiple computing resources at regular intervals.

In one embodiment, the query request includes the target privacy algorithm adopted for the target privacy computing task, and the resource information includes the privacy algorithm supported by the computing resource. Correspondingly, in this step, the management node may determine a number of computing resources supporting the target privacy algorithm from a plurality of computing resources to form the above-mentioned resource list. In a specific embodiment, the resource information also includes the current usage status of the computing resources, and accordingly, this step may further include: including the current usage status information of the above-mentioned several computing resources in the above-mentioned resource list. In another specific embodiment, the query request indicates that the demander for computing resources entrusts the management node to determine the target resource. Correspondingly, this step may also include: the management node, based on the current usage status information of several computing resources, selects the Computing resources whose current use status is idle are selected from the list to form the above resource list, so that the computing resource usage demand side can directly determine the computing resources in the resource list as target resources.

In one embodiment, the above-mentioned target privacy algorithm is not included in the query request. Correspondingly, in this step, the management node can determine some computing resources that are currently idle among the above-mentioned multiple computing resources, and then use each of the several computing resources to The resource ID of the resource and the supported privacy algorithms form a resource list. In a specific embodiment, the idle state may be a completely idle state (not used by any party), or a partially idle state (partial space is used, and the remaining space has not been used). In a specific embodiment, the resource list may also include usage details of several computing resources.

In an embodiment, the above resource list may also include addressing information (such as IP address or domain name) of each computing resource, which is used to subsequently locate the corresponding computing resource and then use the resource. In an embodiment, the above resource list may further include resource IDs of respective computing resources therein.

In this way, the management node can determine the resource list according to the received query request and the stored resource information. Furthermore, in step S230, the management node sends the resource list to the computing resource usage demander. Correspondingly, the computing resource usage demander may receive the resource list, and then in step S240, the computing resource usage demander determines the target resource for executing the target privacy computing task based on the resource list.

In one embodiment, the query request includes a target privacy algorithm adopted for the target privacy computing task, and each computing resource in the correspondingly determined resource list supports the target privacy algorithm. Based on this, in a specific embodiment, this step may include: the demander for computing resource use randomly selects a computing resource from the resource list as the above-mentioned target resource. In another specific embodiment, the resource list also includes the current usage status of each of the computing resources, and correspondingly, the computing resource that is currently in an idle state can be selected as the above-mentioned target resource. In another specific embodiment, the resource list also includes the current usage details of each computing resource, such as how many merchants or data nodes the computing resource is used by, and the CPU, memory, disk, network, etc. of the computing resource in the past one minute. Corresponding to the occupancy situation, the corresponding computing resource can be scored based on the usage details, and the computing resource with the highest score can be determined as the target resource. In an example, usage details of each computing resource may be input into a pre-trained resource scoring model to obtain a corresponding resource score. In another example, resource scoring based on usage details can be combined with pre-specified scoring rules.

In one embodiment, the above resource list includes the privacy algorithms supported by each computing resource. Correspondingly, this step may include: based on the above target privacy algorithm, the computing resource usage demander selects from the above resource list to support the target privacy algorithm Computing resources of , as the target resource above. Further, in a specific embodiment, the computing resources in the resource list are all in an idle state, which means they can be put into use immediately. In another specific embodiment, the computing resources in the resource list are in use state or idle state. If the target resource determined by the computing resource usage demander is in use state, it can reserve the use of the target resource with the management node.

In one embodiment, the computing resource usage demander entrusts the management node to select the target resource in advance, and accordingly, the computing resource usage demander can directly determine the computing resource in the resource list as the target resource.

In this way, the computing resource usage demander can determine the target resource for performing the target privacy task.

According to another embodiment, referring to FIG. 5, the above-mentioned multiple parties may also include other participants (for example, bank B shown in FIG. Other parties that participate in the execution of the above-mentioned target computing tasks. In an example, it is assumed that the demander for computing resource usage is a certain data provider, and correspondingly, other participants may include other data providers.

As shown in FIG. 5, after step S240, the above interaction process may further include step S245 and step S246. In step S245, the computing resource usage requesting party sends the addressing information (such as IP address or domain name) of the target resource to other participating parties. In a specific embodiment, the public key of the target resource is also sent to other participants, so as to facilitate subsequent encrypted transmission of data or digital signature verification of the target resource. In step S246, the computing resource usage demander and other participants jointly use the above target resources to execute the above target privacy computing tasks. In an example, assuming that the computing resource issuer shown in FIG. 5 publishes the above-mentioned target resource, the computing resource issuer may participate in the execution of the target privacy computing task as a participant.

According to yet another embodiment, after step S240, the above-mentioned interaction process may further include: sending an authorization use request for the target resource to the management node from the computing resource usage demand direction; the management node generates a corresponding authorization token according to the authorization use request, And send it to the demander for computing resource use; the demander for computing resource use sends the authorization token to the above-mentioned other participants, after that, the demander for computing resource use and other participants each send resources for the target resource to the publisher of the target resource Use request, the resource use request includes an authorization token, and after the issuer confirms that the authorization token has passed the verification, open the use permission of the target resource to the computing resource usage demander and other participants, allowing them to access and use the target resource.

According to yet another embodiment, after step S240, the above interaction process may further include: sending an authorized use request for the target resource to the management node from the computing resource usage request, and the authorized use request includes identification information of other participants; The node generates a first authorization token according to the identification of the computing resource usage demander, and sends the first authorization token to the computing resource usage demander, and the management node generates a third authorization token according to the identification of other participants, and Send the third authorization token to other participants; after that, the computing resource use demander and other participants each send a resource usage request for the target resource to the issuer of the target resource, and the resource usage request includes the authorization of the corresponding participant Token, after the issuer confirms that the authorization token has passed the verification, open the usage authority of the target resource to the computing resource usage demander and other participants, allowing them to access and use the target resource. In this way, the management node issues authorization tokens to different participants, which can prevent the authorization token from being stolen and cause the target resource to be illegally occupied.

In an example, it is assumed that the above-mentioned target privacy computing task is a privacy data statistics task, the computing resource usage demander and other participants in the above-mentioned multiple parties are data party A and data party B respectively, and the determined target resources are those that provide SGX DB services SGX server, based on this, Figure 6 shows a schematic diagram of the architecture of performing private data statistics tasks according to an embodiment, data party A and data party B are based on their installed SGX client, respectively use the public key of the SGX server to hold Encrypt the private data, and send the encrypted data to the SGX server; then, the data party A sends a SQL query statement to the SGX server to query the statistical indicators of the private data (such as the average annual expenditure of the user), at this time, the SGX server The private key corresponding to the above public key can be used to decrypt the encrypted data sent by data party A and data party B respectively, and calculate the index value of the statistical indicator based on the obtained decrypted data, and then return the index value to the data party as the query result a. In this way, the execution of the target privacy computing task can be completed through multi-party collaboration.

According to yet another embodiment, after step S240, the above-mentioned interaction process may further include: sending a first notification to the management node from the computing resource usage demander, which indicates that the above-mentioned target resource is used by the computing resource usage demander; the management node based on the The first notification is to modify the current usage state of the target resource in the resource information, for example, to be in use. Further, in a specific embodiment, the subsequent step may also include: sending a second notification to the management node instructing the computing resource usage demander to stop using the target resource, or instructing the computing resource usage demander to stop using the target resource. Release the occupancy of the target resource. Therefore, in a more specific embodiment, based on the second notification, the management node modifies the current usage status of the target resource in the resource information, for example, modifies it to be idle. In another more specific embodiment, based on the first notification and the second notification, the management node calculates the cost generated by the computing resource usage requester for using the target resource. In an example, both the first notification and the second notification include the resource ID of the target resource, so that the management node can quickly query the target resource. In one example, the first notification and the second notification each include a time stamp, and correspondingly, the usage fee for the target resource can be calculated according to the calculated usage duration and the preset usage unit price per unit duration. In another example, the first notification also includes information such as the memory space used by the target resource and the number of computing chips (such as CPU, GPU, FPGA, etc.), and accordingly, the usage fee can be calculated in combination with these specific usage information . In another example, the management node can obtain the usage log of the target resource, and by analyzing the usage log, determine the usage information of the target resource by the above-mentioned computing resource usage demander and other participants, and then according to the usage information and preset Calculate the usage fee according to the specified charging standard. Further, in an example, the first notification and/or the second notification may include identification information of other participants participating in the task execution, so that the management node can correspondingly extract the usage information from the usage log.

As a result, the management node can calculate the cost of using the above-mentioned target resources in the process of performing the above-mentioned target privacy computing tasks, and then charge the fee to the computing resource usage demander, or charge the computing resource usage demander and other participants Fees, for example, can be divided equally according to the total number of participants of computing resource usage demanders and other participants, and then charged to each participant according to the equally divided fees. Furthermore, the management node can draw a part of the total fee collected as a handling fee, and transfer the remaining part to the account of the publisher of the target resource. In this way, it is possible to realize the mutual benefits of computing resource usage demanders, publishers and management nodes, and win-win cooperation.

To sum up, in the method for determining computing resources in privacy computing disclosed in the embodiment of this specification, the management node supports computing resource holders to register their computing resources, thereby establishing a resource pool. The demander of the demand can send a resource query request to the management node, and based on the resource list returned by the management node, determine the target resource used to perform the target privacy computing task, and then the computing resource usage demander can cooperate with other participants to jointly use The target resource completes the execution of the target privacy computing task. In this way, by introducing management nodes to establish and manage computing resource pools, the holders and demanders of computing resources can release and discover resources conveniently and quickly.

Corresponding to the above determination method, the embodiment of this specification also discloses a determination device and a determination system.

Fig. 7 shows a schematic structural diagram of an apparatus for determining computing resources in privacy computing according to an embodiment, and the apparatus is integrated in a management node. As shown in Figure 7, the device 700 includes:

The query request receiving unit 710 is configured to receive a query request for computing resources from a computing resource usage demander, wherein the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered at the management node; the resource list determining unit 720 , configured to determine a resource list based on the query request and the resource information of a plurality of locally registered computing resources; the resource list sending unit 730 is configured to send the resource list to the computing resource usage demander, so that the The computing resource usage demander determines the target resource from the resource list to execute the target privacy computing task.

In one embodiment, the query request includes the target privacy algorithm adopted for the target privacy computing task, and the resource information includes the privacy algorithm supported by the computing resource; the resource list determining unit 720 is specifically configured to: determine Several computing resources supporting the target privacy algorithm form the resource list.

In a specific embodiment, the resource list determining unit 720 is further configured to: include the current usage status information of the several computing resources in the resource list.

In one embodiment, the resource information includes privacy algorithms supported by computing resources; the resource list determining unit 720 is specifically configured to: determine several computing resources that are currently idle among the multiple computing resources; according to the A plurality of computing resources and privacy algorithms supported by each computing resource determine the resource list.

In one embodiment, the apparatus further includes: a first notification receiving unit 742 configured to receive a first notification from the computing resource usage demander, which indicates that the target resource is used by the computing resource usage demander; The first status updating unit 744 is configured to update the usage status of the target resource in the resource information according to the first notification.

In a specific embodiment, the apparatus 700 further includes: a second notification receiving unit 752 configured to receive a second notification from the computing resource usage demander, which instructs the computing resource usage demander to stop using the The target resource; the second status update unit 754 is configured to update the usage status of the target resource in the resource information according to the second notification; the fee calculation unit 756 is configured to calculate based on the first notification and the second notification The computing resources use the costs generated by the demand side for using the target resources.

In one embodiment, the apparatus 700 further includes: a registration request receiving unit 762 configured to receive a registration request from the computing resource holder, which includes registration information for the computing resource, and the registration information includes the information supported by the corresponding computing resource. Privacy algorithm; registration information storage unit 764, configured to include the registration information into the resource information when the registration request is approved.

In a specific embodiment, the registration information storage unit 764 is specifically configured to: allocate a resource ID to the computing resource corresponding to the registration request; and associate and store the resource ID and registration information.

In one embodiment, the registration information further includes at least one of the following: IP address, domain name, and public key of the corresponding computing resource.

Fig. 8 shows a schematic structural diagram of an apparatus for determining computing resources in privacy computing according to another embodiment, and the apparatus is integrated in a demand side for computing resource usage. As shown in Figure 8, the device 800 includes:

The query request sending unit 810 is configured to send a query request for computing resources to the management node, where the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered at the management node; the resource list receiving unit 820 is configured to Receive a resource list from the management node, the resource list is determined based on the query request and the resource information of a plurality of locally registered computing resources; the target resource determining unit 830 is configured to determine an execution target based on the resource list Target resource for private computing tasks.

In one embodiment, the query request sending unit 810 is further configured to: determine the target privacy algorithm adopted for the target privacy computing task, and include the target privacy algorithm in the query request.

In a specific embodiment, the target resource determining unit 830 is specifically configured to: randomly select a computing resource from the resource list as the target resource.

In a specific embodiment, the resource list includes current usage status information of each computing resource; the target resource determining unit 830 is specifically configured to: select a computing resource whose current usage status is idle from the resource list, as The target resource.

In one embodiment, the resource list includes privacy algorithms supported by each computing resource; the target resource determining unit 830 is specifically configured to: based on the target privacy algorithm adopted for the target privacy computing task, select from the resource list A computing resource that supports the target privacy algorithm is selected as the target resource.

In one embodiment, the apparatus 800 further includes: an IP address sending unit 840, configured to send the IP address of the target resource to other participants; a privacy task execution unit 850, configured to utilize the target resource, and The other participants jointly execute the target privacy computing task.

The embodiment of this specification also discloses a system for determining computing resources in privacy computing. The determination system includes:

Computing resource use demand side, used to send a query request for computing resources to the management node, wherein the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered with the management node; the management node is used to The query request and resource information of a plurality of computing resources registered locally, determining a resource list, and sending the resource list to the computing resource usage demander; the computing resource usage demander is configured to list, identifying the target resource for performing the target private computing task.

In one embodiment, the computing resource usage requester is further configured to: determine the target privacy algorithm adopted for the target privacy computing task, and include the target privacy algorithm in the query request; the management node is specifically used to : Determine a number of computing resources supporting the target privacy algorithm, and form the resource list.

In a specific embodiment, the management node is further configured to: include the current usage status information of the several computing resources in the resource list; the computing resource usage demander is specifically configured to: A computing resource whose current use state is idle is selected from the list as the target resource.

In another specific embodiment, the management node is specifically configured to: select a computing resource whose current usage state is idle from the several computing resources based on the current usage state information of the several computing resources, and form the resource list; the computing resource use demander is specifically configured to: determine the computing resource in the resource list as the target resource.

In one embodiment, the management node is specifically configured to: determine several computing resources that are currently idle among the multiple computing resources, and form the resource list; the computing resource usage demander is specifically configured to: based on For the target privacy algorithm adopted by the target privacy computing task, computing resources supporting the target privacy algorithm are selected from the resource list as the target resource.

According to another embodiment, there is also provided a computer-readable storage medium on which a computer program is stored. When the computer program is executed in a computer, the computer is instructed to execute the method described in conjunction with FIG. 2 or FIG. 5 .

According to yet another embodiment, there is also provided a computing device, including a memory and a processor, wherein executable code is stored in the memory, and when the processor executes the executable code, the implementation in conjunction with FIG. 2 or FIG. 5 is realized. the method described.

Those skilled in the art should be aware that, in the above one or more examples, the functions described in the present invention may be implemented by hardware, software, firmware or any combination thereof. When implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. Protection scope, any modification, equivalent replacement, improvement, etc. made on the basis of the technical solution of the present invention shall be included in the protection scope of the present invention.

Claims

A method for determining computing resources in privacy computing, the method is executed by a management node, and the method includes:

Receive a query request for computing resources from a computing resource usage demander, where the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered at the management node;

determining a resource list based on the query request and resource information of a plurality of locally registered computing resources;

Sending the resource list to the computing resource usage demander, so that the computing resource usage demander determines a target resource from the resource list for executing a target privacy computing task.
The method according to claim 1, wherein the query request includes a target privacy algorithm adopted for the target privacy computing task, and the resource information includes a privacy algorithm supported by computing resources; wherein, based on the query request and the resource information of multiple computing resources registered locally to determine the resource list, including:

A number of computing resources supporting the target privacy algorithm are determined to form the resource list.
The method of claim 2, further comprising:

Include the current usage state information of the several computing resources in the resource list.
The method according to claim 1, wherein the resource information includes a privacy algorithm supported by computing resources; wherein, based on the query request and resource information of a plurality of locally registered computing resources, determining a resource list includes:

determining a number of computing resources that are currently in an idle state among the plurality of computing resources;

The resource list is determined according to the plurality of computing resources and the privacy algorithm supported by each of the computing resources.
The method according to claim 1, wherein, after sending the resource list to the computing resource usage demander, the method further comprises:

receiving a first notification from the computing resource usage demander indicating that the target resource is used by the computing resource usage demander;

According to the first notification, update the usage status of the target resource in the resource information.
The method according to claim 5, wherein, after receiving the first notification from the computing resource usage demander, the method further comprises:

receiving a second notification from the computing resource usage demander, which instructs the computing resource usage demander to stop using the target resource;

updating the usage status of the target resource in the resource information according to the second notification;

Based on the first notification and the second notification, calculate the cost generated by the computing resource usage demand side for using the target resource.
The method according to any one of claims 1-6, wherein the resource list includes the IP addresses of each computing resource therein.
The method according to claim 1, wherein, before receiving a query request for computing resources from a computing resource usage demander, the method further comprises:

Receive a registration request from a computing resource holder, including registration information for the computing resource, where the registration information includes a privacy algorithm supported by the corresponding computing resource;

In the case that the registration request is approved, the registration information is included in the resource information.
The method according to claim 8, wherein categorizing the registration information into the resource information comprises:

Allocating a resource ID to the computing resource corresponding to the registration request;

The resource ID and the registration information are associated and stored.
The method according to claim 8 or 9, wherein the registration information further includes at least one of the following: IP address, domain name, and public key of the corresponding computing resource.
A method for determining computing resources in privacy computing, the method is executed by a computing resource usage demander, and the method includes:

Send a query request for computing resources to the management node, where the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered with the management node;

receiving a resource list from the management node, the resource list being determined based on the query request and resource information of a plurality of locally registered computing resources;

Based on the resource list, a target resource for executing a target privacy computing task is determined.
The method according to claim 11, wherein the method further comprises:

Determine the target privacy algorithm adopted for the target privacy computing task, and include the target privacy algorithm in the query request.
The method according to claim 12, wherein the resource list includes current usage status information of each computing resource; wherein, based on the resource list, determining a target resource for executing a target privacy computing task comprises:

Selecting a computing resource that is currently in an idle state from the resource list as the target resource.
The method according to claim 11, wherein the resource list includes privacy algorithms supported by each computing resource; wherein, based on the resource list, determining a target resource for executing a target privacy computing task includes:

Based on the target privacy algorithm adopted for the target privacy computing task, a computing resource supporting the target privacy algorithm is selected from the resource list as the target resource.
The method according to claim 11, wherein, after determining the target resource for performing the target privacy computing task, the method further comprises:

sending the IP address of the target resource to other participants;

Using the target resource, jointly execute the target privacy computing task with the other participants.
A method for determining computing resources in privacy computing, comprising:

Sending a query request for computing resources to the management node on demand for computing resources, where the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered with the management node;

The management node determines a resource list based on the query request and resource information of a plurality of computing resources locally registered by it;

The management node sends the resource list to the computing resource usage demander;

Based on the resource list, the computing resource usage demander determines the target resource for executing the target privacy computing task.
The method according to claim 16, wherein, the method further comprises: the computing resource utilization requesting party determining the target privacy algorithm adopted for the target privacy computing task, and including the target privacy algorithm in the query request;

Wherein, the management node determines a resource list based on the query request and resource information of multiple computing resources stored therein, including: determining several computing resources supporting the target privacy algorithm to form the resource list.
The method according to claim 17, wherein the method further comprises: the management node includes current usage status information of the several computing resources in the resource list;

Wherein, the computing resource use requesting party determines the target resource for executing the target privacy computing task based on the resource list, including: selecting a computing resource that is currently in an idle state from the resource list as the target resource.
The method according to claim 17, wherein determining several computing resources that support the target privacy algorithm and forming the resource list includes: the management node, based on the current usage status information of the several computing resources, from the several Selecting computing resources whose current use status is idle from the computing resources to form the resource list;

Wherein, the computing resource usage demander determines the target resource for executing the target privacy computing task based on the resource list, including: the computing resource usage demander determines the computing resource in the resource list as the target resource .
The method according to claim 17, wherein the management node determines the resource list based on the query request and the resource information of multiple computing resources stored therein, comprising: determining that the multiple computing resources are currently in an idle state A number of computing resources to form the resource list;

Wherein, based on the resource list, determining the target resource for executing the target privacy computing task includes: based on the target privacy algorithm adopted for the target privacy computing task, selecting computing resources supporting the target privacy algorithm from the resource list , as the target resource.
A device for determining computing resources in privacy computing, the device is integrated in a management node, and the device includes:

The query request receiving unit is configured to receive a query request for computing resources from a demander for computing resources, where the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered at the management node;

A resource list determining unit configured to determine a resource list based on the query request and resource information of a plurality of locally registered computing resources;

A resource list sending unit configured to send the resource list to the computing resource usage demander, so that the computing resource usage demander can determine target resources from the resource list for executing target privacy computing tasks.
A device for determining computing resources in privacy computing, the device is integrated with a computing resource usage demand side, and the device includes:

A query request sending unit configured to send a query request for computing resources to a management node, where the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered with the management node;

A resource list receiving unit configured to receive a resource list from the management node, the resource list is determined based on the query request and resource information of a plurality of locally registered computing resources;

The target resource determining unit is configured to determine a target resource for executing a target privacy computing task based on the resource list.
A system for determining computing resources in privacy computing, including:

The demander for computing resource use is used to send a query request for computing resources to the management node, where the computing resources are public resources that support privacy algorithms, and the computing resources are pre-registered with the management node;

The management node is configured to determine a resource list based on the query request and resource information of a plurality of computing resources locally registered by it, and send the resource list to the computing resource usage demander;

The computing resource usage demand side is further configured to determine target resources for executing target privacy computing tasks based on the resource list.
A computer-readable storage medium, on which a computer program is stored, wherein, when the computer program is executed in a computer, the computer is caused to execute the method described in any one of claims 1-20.
A computing device, comprising a memory and a processor, wherein executable code is stored in the memory, and the method according to any one of claims 1-20 is implemented when the processor executes the executable code.