Method and device for verification and configuration of multi-party prediction services in institutions

Abstract

The embodiments of this specification provide a method and apparatus for verifying and configuring a multi-party prediction service in an institution. The verification method includes: obtaining a sample object set used by a target institution for training a local prediction model in a multi-party secure computation and the training results corresponding to the sample object set. The sample object set is used as a mock test case for the target institution to make predictions based on a virtual link to test the prediction results of the target institution for the sample object set. The virtual link is a link for providing prediction services constructed by the target institution after completing the multi-party secure computation training of the local prediction model. The virtual link is verified based on the degree of match between the training results of the sample object set and the prediction results.

Description

Technical Field

[0001] This document belongs to the field of artificial intelligence technology, and more particularly to a method and apparatus for institutional verification and configuration of a multi-party prediction service. Background Art

[0002] Multi-party secure computation (MPCC) is a machine learning framework that enables joint modeling across multiple organizations while adhering to user privacy and regulatory requirements. Member organizations must exchange local private data for training according to the MPC protocol. Under the MPC protocol, private data is encrypted and desensitized to maintain privacy. Even after desensitization, the desensitized data retains its mathematical validity as plaintext, without impacting model training.

[0003] Currently, some applications based on multi-party secure computation deploy interfaces within various institutions to provide prediction services. Upon receiving a client's prediction request through a local interface, any institution will leverage models trained locally and across other institutions to provide a joint prediction. This process involves multiple steps, and any issues that arise could result in service anomalies. Therefore, ensuring the correct implementation of these multi-party prediction services within institutions is a pressing technical challenge. Summary of the Invention

[0004] The embodiment of this specification aims to provide a method and apparatus for verifying and configuring a multi-party prediction service in an institution, which can ensure that the multi-party prediction service is correctly put into use on the institution side.

[0005] In order to achieve the above purpose, the embodiments of this specification are implemented as follows:

[0006] First, a verification method for a multi-party prediction service in an institution is provided, including:

[0007] Obtaining a sample object set used by the target organization for training a local prediction model in a multi-party secure computation and a training result corresponding to the sample object set;

[0008] Using the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model.

[0009] The virtual link is verified based on a matching degree between a training result and a prediction result of the sample object set.

[0010] Secondly, a method for configuring a multi-party prediction service in an institution is provided, including:

[0011] Before configuring an actual link for providing prediction services to a target organization, obtaining a sample object set used by the target organization for training a local prediction model in the multi-party secure computation and a training result corresponding to the sample object set;

[0012] Using the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model.

[0013] Verifying the virtual link based on a degree of matching between a training result and a prediction result of the sample object set;

[0014] When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

[0015] In a third aspect, a verification device for a multi-party prediction service in an institution is provided, comprising:

[0016] A multi-party secure computing tracing module obtains a sample object set used by the target institution to train a local prediction model in a multi-party secure computing and the training results corresponding to the sample object set;

[0017] A mock testing module, which uses the sample object set as a mock test case for the target organization to perform prediction based on the virtual link, so as to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization after completing multi-party secure computation training on a local prediction model to provide prediction services;

[0018] The virtual link verification module verifies the virtual link based on the matching degree between the training result and the prediction result of the sample object set.

[0019] In a fourth aspect, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is executed by the processor:

[0020] Obtaining a sample object set used by the target organization for training a local prediction model in a multi-party secure computation and a training result corresponding to the sample object set;

[0021] Using the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model.

[0022] The virtual link is verified based on a matching degree between a training result and a prediction result of the sample object set.

[0023] In a fifth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0024] Obtaining a sample object set used by the target organization for training a local prediction model in a multi-party secure computation and a training result corresponding to the sample object set;

[0025] Using the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model.

[0026] The virtual link is verified based on a matching degree between a training result and a prediction result of the sample object set.

[0027] In a sixth aspect, a device for configuring a multi-party prediction service in an organization is provided, comprising:

[0028] The multi-party secure computing tracing module obtains the sample object set used by the target organization to train the local prediction model in the multi-party secure computing and the training results corresponding to the sample object set before configuring the actual link for providing prediction services based on the local prediction model to the target organization;

[0029] A mock testing module uses the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, so as to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization after completing multi-party secure computation training on a local prediction model to simulate the actual link after configuration to provide prediction services;

[0030] The virtual link verification module verifies the virtual link based on the matching degree between the training result and the prediction result of the sample object set. The actual link configuration module configures the actual link of the target organization to provide prediction service according to the virtual link when the virtual link passes the verification.

[0031] In a seventh aspect, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is executed by the processor:

[0032] Before configuring an actual link for providing prediction services based on the local prediction model to the target organization, obtaining a sample object set used by the target organization for training the local prediction model in the multi-party secure computation and a training result corresponding to the sample object set;

[0033] Using the sample object set as a mock test case for the target organization's prediction based on a virtual link to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization after completing multi-party secure computation training on a local prediction model to simulate the actual link after configuration to provide prediction services;

[0034] Verifying the virtual link based on a degree of matching between a training result and a prediction result of the sample object set;

[0035] When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

[0036] In an eighth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0037] Before configuring an actual link for providing prediction services based on the local prediction model to the target organization, obtaining a sample object set used by the target organization for training the local prediction model in the multi-party secure computation and a training result corresponding to the sample object set;

[0038] Using the sample object set as a mock test case for the target organization's prediction based on a virtual link to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization after completing multi-party secure computation training on a local prediction model to simulate the actual link after configuration to provide prediction services;

[0039] Verifying the virtual link based on a degree of matching between a training result and a prediction result of the sample object set;

[0040] When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

[0041] The solution of the embodiments of this specification constructs a virtual link on the institution side that provides a multi-party prediction service based on a prediction model of multi-party secure computation, and uses the sample objects used by the institution for training during multi-party secure computation to perform mock testing on the virtual link, thereby verifying the logic of the virtual link based on the matching degree between the prediction results corresponding to the sample objects during the mock test and the corresponding training results during the multi-party secure computation. The verification results of the virtual link can be used to ensure that the multi-party prediction service can be put into normal use on the institution side. For example, before configuring the actual link of the multi-party prediction service on the institution side, the constructed virtual link is first mock tested, and after the mock test passes, the actual link of the multi-party prediction service is configured. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this specification. For ordinary relevant personnel in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0043] Figure 1 This is a flowchart of a verification method for a multi-party prediction service in an institution according to an embodiment of this specification.

[0044] Figure 2 This is a schematic diagram of the mock test of the virtual link implemented in this specification.

[0045] Figure 3 This is a flowchart of a method for configuring a multi-party prediction service in an organization according to an embodiment of this specification.

[0046] Figure 4 This is a schematic diagram of the structure of the verification device of the multi-party prediction service in the institution according to the embodiment of this specification.

[0047] Figure 5 This is a structural diagram of the configuration device of the multi-party prediction service in the organization according to the embodiment of this specification.

[0048] Figure 6 Schematic diagram of the structure of an electronic device according to an embodiment of this specification. DETAILED DESCRIPTION

[0049] In order to help those skilled in the art better understand the technical solutions in this specification, the following will clearly and completely describe the technical solutions in the embodiments of this specification in conjunction with the drawings in the embodiments of this specification. Obviously, the embodiments described are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary relevant persons in this field without making creative efforts should fall within the scope of protection of this specification.

[0050] As mentioned previously, some current applications based on multi-party secure computation rely on interfaces deployed by various institutions to provide prediction services. Upon receiving a prediction request from a client through a local interface, any institution will call upon trained models locally and across other institutions to perform a joint prediction. This process involves multiple steps, and any issues that arise can lead to service anomalies. Therefore, this document proposes a technical solution to ensure the correct implementation of multi-prediction service links on the institution side.

[0051] Figure 1 This is a flowchart of a verification method for a multi-party prediction service in an institution according to an embodiment of the present specification. Figure 1 The method shown can be performed by the corresponding device below, and specifically includes the following steps:

[0052] S102, obtaining a sample object set used by the target organization for training a local prediction model in a multi-party secure computation and a training result corresponding to the sample object set.

[0053] In the embodiments of this specification, the various organizations participating in the multi-party secure computation jointly provide prediction services. That is, upon receiving a prediction request from a local client, any target organization will invoke prediction models from both the local and other organizations to perform predictions. These prediction results are ultimately aggregated by the target organization to generate a final prediction result that is fed back to the local client.

[0054] During the training process of multi-party secure computing, each institution, based on the multi-party secure computing protocol, desensitizes a portion of local training data and provides it to the multi-party secure computing platform, so that the multi-party secure computing platform can use the training data provided by each institution to train the prediction model. After the training is completed, the multi-party secure computing platform will feed back the obtained model parameters to each institution, so that each institution can adjust the prediction model trained locally with the model parameters provided by the multi-party secure computing platform as a reference. Alternatively, each institution can also desensitize at least a portion of the local training data based on the multi-party secure computing protocol (the decryption method can be but is not limited to: secret sharing, oblivious transmission, obfuscated circuits, homomorphic encryption, zero-knowledge proof, etc.) and send it to other institutions, and complete the training of the prediction model using only its own computing power and the training data sent by local and other institutions.

[0055] It should be understood that, except for the training data shared between institutions through the multi-party secure computation protocol, the remaining training data is not publicly available and cannot be guaranteed to be identical. Therefore, the prediction biases of each institution's locally trained prediction models may differ, and the prediction results for the same object may differ after deployment. It is precisely through the joint prediction of multiple institutions' prediction models that this prediction analysis can be more comprehensive, thereby improving prediction accuracy.

[0056] The sample objects in the sample object set in this step specifically refer to the shared sample objects determined by the target organization and the other organizations based on the intersection of the privacy-preserving set. Normally, using the feature data of the shared sample objects across the organizations to invoke a virtual link for prediction, the prediction results should be consistent with the training results. Therefore, the virtual link can be verified based on the degree of match between the prediction results and the training results.

[0057] Among them, privacy-preserving set intersection can ensure that institutions can determine common sample objects without revealing their own sample object information. Since this is an existing technology, it will not be further explained in this article.

[0058] S104, using the sample object set as a mock test case of the target organization based on a virtual link to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computing training on the local prediction model.

[0059] After the multi-party secure computation training process is complete, each institution needs to build its own chain to provide prediction services. This involves invoking the prediction models of other institutions to perform predictions on the target data. For ease of understanding, let's take the example of providing transaction risk prediction services, where banks collaborate with internet finance platforms. When a local user of an internet finance platform requests a transaction risk prediction from the platform, the platform not only uses its own risk prediction model but also requests the bank's risk prediction model to provide a risk prediction. This process involves cross-system interaction between the internet finance platform and the bank. When establishing local risk prediction service chains, internet finance platforms must ensure the correctness of cross-platform interaction with banks.

[0060] To this end, this step establishes a virtual link for verification before configuring the actual link for the target institution to provide the forecasting service. For example, the internet finance platform can build a virtual bank object based on the bank's system and establish the execution logic required for joint forecasting with the virtual bank object. This includes logic for sending the forecasted data received locally by the internet finance platform to the virtual bank object, logic for the virtual bank object to use a local forecasting model to predict the forecasted data, and logic for the virtual bank object to feed the locally generated forecast results back to the internet finance platform.

[0061] It should be noted that the virtual link in the embodiment of this specification is mainly used to verify the correctness of the cross-system interaction between the target organization and other organizations, and does not need to consider the model performance of the prediction model of other organizations. Therefore, the prediction model of the target organization can be used as the prediction model of other organizations in the virtual link to ensure that the virtual link can operate successfully. There is no need for other organizations to provide their own prediction models to the target organization for support.

[0062] It should also be understood that the actual construction of a virtual link involves many factors, such as message encapsulation, information expansion, and identity authentication. Problems in any of these links may cause errors in the joint prediction results. By verifying the virtual link, these problems can be discovered in a timely manner before the actual link configuration.

[0063] In an embodiment of this specification, a data management system may be deployed in the target organization to record and maintain sample object sets and feature data corresponding to the sample object sets used by the target organization to train local prediction models in multi-party secure computing.

[0064] Correspondingly, this step obtains sample objects (shared sample objects) and corresponding feature data from the target organization's data management system, and requests the feature data of the sample objects to the target organization's virtual link to complete the corresponding prediction.

[0065] S106 , verifying the virtual link based on the matching degree between the training result and the prediction result of the sample object set.

[0066] It should be understood that if the virtual link is normal, the training results of any sample object in the sample object set will be consistent with the predicted results. Therefore, the proportion of sample objects whose training results in the sample object set (the training results corresponding to the last round of multi-party security calculation) match the predicted results can be used as the basis for judging whether the virtual link is normal.

[0067] For example, if the prediction model used by another institution in a virtual link is replaced by the target institution's prediction model, or if the prediction model provided by the other institution is unmodified after training, the virtual link can be considered verified if the proportion of sample objects in the sample set whose training results are consistent with the prediction results reaches a preset threshold of 99%. In other words, if the same prediction model, or a prediction model trained using the same sample objects, is used to simulate joint predictions of shared sample objects, theoretically there should be no deviation between the prediction results of the other institution and the local training results. Therefore, the preset threshold should be set at a relatively high level.

[0068] Correspondingly, if the prediction model of other institutions used by the virtual link is a real prediction model provided by the other institutions in real time, it is appropriate to allow a certain degree of deviation between the prediction results of other institutions and the local training results of the target institution (other institutions may adjust their own prediction models according to their own needs, resulting in certain differences between the predictions for common sample objects and the results of the original training). For example, when the proportion of sample objects in the sample object set whose training results are consistent with the prediction results reaches 95% (slightly lower than the 99% introduced above), the virtual link can be judged to have passed the verification.

[0069] The verification method of the embodiments of this specification constructs a virtual link on the institution side that provides a multi-party prediction service based on a prediction model of multi-party secure computation, and uses the sample objects used by the institution for training during multi-party secure computation to perform a mock test on the virtual link, thereby verifying the logic of the virtual link based on the matching degree between the prediction results corresponding to the sample objects during the mock test and the corresponding training results during the multi-party secure computation. The verification results of the virtual link can be used to ensure that the multi-party prediction service can be put into normal use on the institution side. For example, before configuring the actual link of the multi-party prediction service on the institution side, the constructed virtual link is first mock tested, and after the mock test passes, the actual link of the multi-party prediction service is configured. Figure 2 This is a flowchart for mock testing a virtual link. The virtual link for the target organization to provide prediction services includes two nodes: the target organization and another organization (the virtual link can also include nodes from more organizations, which will not be detailed here).

[0070] The target organization and other organizations are each equipped with their own data management systems and prediction models. The data management system records the feature data of the sample object sets used by the affiliated organization during the multi-party secure computation training process. When the mock test program runs, it retrieves the sample object sets shared by the target organization and other organizations from the target organization's data management system. It then initiates mock test requests to the target organization for each sample object in the sample object set, calling the target organization and inputting the feature data of the sample object set recorded in the local data management system into the local prediction model for prediction, thereby obtaining the target organization's prediction results.

[0071] Furthermore, when the mock test program is running, it simulates the target organization forwarding the mock test request to other organizations, instructing them to input the feature data of the sample object set recorded in their local data management system into their local prediction model for prediction, thereby obtaining the prediction results of the other organizations. Simultaneously, the other organizations send their local prediction results to the target organization, which then performs a weighted calculation based on the local prediction results and the prediction results of the other organizations to obtain a final prediction result, which is then fed back to the mock test program.

[0072] The mock test program then sends the final prediction results for the sample object set to the data processing platform, which calculates the degree of match between the training results and the prediction results for each sample object in the sample object set. If the percentage of sample objects whose training results match the prediction results reaches a preset threshold, the data processing platform determines that the virtual link has passed verification; otherwise, it is determined to have failed verification. The data processing platform can obtain the training results corresponding to the sample object set from the data management system of the target organization or other organizations, but this is not specifically limited here.

[0073] Based on the above principles, before configuring the actual link that provides prediction services to the target organization, a virtual link that simulates the changes can be pre-built for mock testing. After the virtual link passes the verification of the mock test, the actual link can be configured according to the virtual link.

[0074] Correspondingly, Figure 3 This is a configuration method of the multi-party prediction service in an organization according to an embodiment of this specification. Figure 3 The method shown can be performed by the corresponding device below, and specifically includes the following steps:

[0075] S302, before configuring an actual link for providing prediction services based on a local prediction model for the target organization, obtain a sample object set used by the target organization for training the local prediction model in a multi-party secure computation and a training result corresponding to the sample object set.

[0076] S304, constructing a mock test case for the target organization to predict the sample object set based on a virtual link, so as to run the mock test case to obtain the prediction result of the sample object set. The virtual link is a link constructed by the target organization after completing the multi-party secure computing training of the local prediction model to simulate the actual link after configuration to provide prediction services.

[0077] S306 , verifying the virtual link based on the matching degree between the training result and the prediction result of the sample object set.

[0078] S308: When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

[0079] The configuration described here can refer to the initial deployment of an actual link for a multi-party prediction service on the organization side, or changes to an actual link for an already deployed multi-party prediction service on the organization side. It should be understood that regardless of how the actual link is modified after deployment, as long as the logic of each organization's prediction model remains unchanged, the feature data of the sample objects used in the mock testing of the virtual link will be the previous model training data, and no adaptive updates are required to accommodate changes to the actual link.

[0080] The configuration method of the embodiment of this specification first constructs a virtual link for simulating the actual link after configuration to provide prediction services for the target organization based on the prediction model of multi-party secure computing, and uses the sample objects used by the organization for training during multi-party secure computing to perform mock testing on the virtual link, thereby verifying the logic of the virtual link based on the matching degree between the prediction results corresponding to the sample objects during the mock test and the corresponding training results during multi-party secure computing, and then configures the actual link after the virtual link passes the verification, thereby ensuring that the multi-party prediction service can be put into use normally on the side of each organization and providing accurate prediction services to the clients of each organization.

[0081] In addition, corresponding to Figure 1 In addition to the method shown, the embodiment of this specification also provides a verification device for a multi-party prediction service in an institution. Figure 4 4 is a schematic diagram of the structure of the verification device 400 according to an embodiment of the present specification, including:

[0082] The multi-party secure computation tracing module 410 obtains the sample object set used by the target institution to train the local prediction model in the multi-party secure computation and the training results corresponding to the sample object set.

[0083] The mock testing module 420 uses the sample object set as a mock test case for the target organization to perform prediction based on the virtual link, so as to test the prediction result of the target organization for the sample object set.

[0084] The virtual link verification module 430 verifies the virtual link based on the matching degree between the training result and the prediction result of the sample object set.

[0085] The verification device of the embodiment of this specification constructs a virtual link on the institution side that provides a multi-party prediction service based on a prediction model of multi-party secure computing, and uses the sample objects used by the institution for training during multi-party secure computing to perform mock testing on the virtual link, thereby verifying the logic of the virtual link based on the matching degree between the prediction results corresponding to the sample objects during the mock test and the corresponding training results during multi-party secure computing. The verification results of the virtual link can be used to ensure that the multi-party prediction service can be put into use normally on the institution side. For example, before configuring the actual link of the multi-party prediction service on the institution side, the constructed virtual link is first mock tested, and after the mock test passes, the actual link of the multi-party prediction service is configured.

[0086] Optionally, the process of the target institution providing the prediction service includes: the target institution predicts the data to be predicted based on a local prediction model and a prediction model called from other institutions of the multi-party secure computing to provide a prediction result.

[0087] The sample objects of the sample object set are common sample objects determined by the target organization and the other organizations based on the intersection of the privacy protection set.

[0088] Optionally, the virtual link passing verification means that the proportion of sample objects in the sample object set whose training results match the prediction results reaches a preset ratio threshold.

[0089] Optionally, after the mock test case is run, the feature data of the sample object set used by the target organization to train the local prediction model is obtained from the data management system of the target organization, and the feature data of the used sample object set is requested to the virtual link of the target organization to complete the prediction for the sample object set.

[0090] Optionally, the training result corresponding to the sample object set is the training result corresponding to the last round of the multi-party secure computation performed by the target organization.

[0091] Obviously, the verification device of the embodiment of this specification can be used as the above Figure 1 The execution subject of the method shown can thus realize the method in Figure 1 The steps and functions implemented are not described in detail in this article because the principles are the same.

[0092] In addition, corresponding to Figure 2In addition to the method shown, the embodiment of this specification also provides a configuration device for a multi-party prediction service in an organization. Figure 5 : is a schematic diagram of the configuration device 500 according to an embodiment of this specification, including:

[0093] The multi-party secure computing tracing module 510 obtains the sample object set used by the target organization to train the local prediction model in the multi-party secure computing and the training results corresponding to the sample object set before configuring the actual link for providing prediction services based on the local prediction model to the target organization.

[0094] The mock testing module 520 uses the sample object set as a mock test case for the target organization to perform prediction based on the virtual link, so as to test the prediction result of the target organization for the sample object set.

[0095] The virtual link verification module 530 verifies the virtual link based on the matching degree between the training result and the prediction result of the sample object set.

[0096] The actual link configuration module 540 configures an actual link for providing prediction services to the target organization according to the virtual link when the virtual link passes the verification.

[0097] Before configuring an actual link for providing prediction services based on a prediction model for multi-party secure computing to a target organization, the configuration device of an embodiment of this specification first constructs a virtual link for simulating the actual link after configuration to provide prediction services, and uses the sample objects used by the organization for training during multi-party secure computing to perform a mock test on the virtual link, thereby verifying the logic of the virtual link based on the matching degree between the prediction results corresponding to the sample objects during the mock test and the corresponding training results during multi-party secure computing, and then configuring the actual link after the virtual link passes the verification, thereby ensuring that the multi-party prediction service can be put into use normally on the side of each organization and providing accurate prediction services to the clients of each organization.

[0098] Optionally, the process of the target institution providing the prediction service includes: the target institution predicts the data to be predicted based on a local prediction model and a prediction model called from other institutions of the multi-party secure computing to provide a prediction result.

[0099] The sample objects of the sample object set are common sample objects determined by the target organization and the other organizations based on the intersection of the privacy protection set.

[0100] Optionally, the virtual link passing verification means that the proportion of sample objects in the sample object set whose training results match the prediction results reaches a preset ratio threshold.

[0101] Optionally, after the mock test case is run, the feature data of the sample object set used by the target organization to train the local prediction model is obtained from the data management system of the target organization, and the feature data of the used sample object set is requested to the virtual link of the target organization to complete the prediction for the sample object set.

[0102] Optionally, the training result corresponding to the sample object set is the training result corresponding to the last round of the multi-party secure computation performed by the target organization.

[0103] Obviously, the configuration device of the embodiment of this specification can be used as the above Figure 3 The execution subject of the method shown can thus realize the method in Figure 3 The steps and functions implemented are not described in detail in this article because the principles are the same.

[0104] Figure 6 This is a schematic diagram of the structure of an electronic device according to an embodiment of this specification. Figure 6 At the hardware level, the electronic device includes a processor and, optionally, an internal bus, a network interface, and memory. The memory may include internal memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for its services.

[0105] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0106] The memory is used to store programs. Specifically, the program may include program code, which includes computer operating instructions. The memory may include internal memory and non-volatile memory, and provides instructions and data to the processor.

[0107] The processor reads the corresponding computer program from the non-volatile memory into the internal memory and then runs it, forming the above-mentioned verification device at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0108] Obtain a sample object set used by the target organization for training a local prediction model in multi-party secure computing and a training result corresponding to the sample object set.

[0109] The sample object set is used as a mock test case for the target organization to perform prediction based on a virtual link to test the prediction results of the target organization for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computing training on the local prediction model.

[0110] The virtual link is verified based on a matching degree between a training result and a prediction result of the sample object set.

[0111] Alternatively, the processor reads the corresponding computer program from the non-volatile memory into the internal memory and then runs it, forming the above-mentioned configuration device at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0112] Before configuring an actual link for providing prediction services based on a local prediction model for a target organization, a sample object set used by the target organization for training the local prediction model in the multi-party secure computation and a training result corresponding to the sample object set are obtained.

[0113] The sample object set is used as a mock test case for the target organization to perform prediction based on a virtual link to test the prediction results of the target organization for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computing training on the local prediction model.

[0114] Verifying the virtual link based on a degree of matching between a training result and a prediction result of the sample object set;

[0115] When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

[0116] The above is as in this manual Figure 1 or Figure 3The methods disclosed in the illustrated embodiments can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in the processor or by software instructions. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of this specification can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in conjunction with the embodiments of this specification can be directly implemented as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0117] It should be understood that the electronic device of the embodiment of this specification can implement the above-mentioned evidence collection method in Figure 1 or Figure 2 The functions of the methods shown are as follows. Since the principles are the same, they will not be described in detail in this article.

[0118] Of course, in addition to software implementation, the electronic device in this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0119] In addition, an embodiment of this specification also proposes a computer-readable storage medium, which stores one or more programs, and the one or more programs include instructions.

[0120] When the instruction is executed by a portable electronic device including multiple applications, the portable electronic device can execute Figure 1 The method of the embodiment shown is specifically used to perform the following steps:

[0121] Obtain a sample object set used by the target organization for training a local prediction model in multi-party secure computing and a training result corresponding to the sample object set.

[0122] The sample object set is used as a mock test case for the target organization to perform prediction based on a virtual link to test the prediction results of the target organization for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computing training on the local prediction model.

[0123] The virtual link is verified based on a matching degree between a training result and a prediction result of the sample object set.

[0124] Alternatively, when the above instruction is executed by a portable electronic device including a plurality of applications, the portable electronic device can execute Figure 3 The method of the embodiment shown is specifically used to perform the following steps:

[0125] Before configuring an actual link for providing prediction services based on a local prediction model for a target organization, a sample object set used by the target organization for training the local prediction model in the multi-party secure computation and a training result corresponding to the sample object set are obtained.

[0126] The sample object set is used as a mock test case for the target organization to perform prediction based on a virtual link to test the prediction results of the target organization for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computing training on the local prediction model.

[0127] Verifying the virtual link based on a degree of matching between a training result and a prediction result of the sample object set;

[0128] When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

[0129] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0130] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0131] The above are merely examples of the present invention and are not intended to limit the present invention. For relevant personnel in this field, this specification may be subject to various modifications and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this specification shall be included in the scope of the claims of this specification. In addition, all other embodiments obtained by ordinary relevant personnel in this field without making any creative work shall fall within the scope of protection of this document.

Claims

1. A method for verifying a multi-party prediction service in an institution, comprising: Obtaining a sample object set used by the target organization for training a local prediction model in a multi-party secure computation and a training result corresponding to the sample object set; The sample objects of the sample object set are common sample objects determined by the target organization and other organizations based on the intersection of the privacy protection set, and the other organizations belong to the multi-party secure computation; Using the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model. The virtual link is verified based on a matching degree between a training result and a prediction result of the sample object set.

2. The method according to claim 1, The process of the target institution providing the forecast service includes: The target institution predicts the data to be predicted based on the local prediction model and the prediction model called from other institutions of the multi-party secure computing to provide a prediction result.

3. The method according to claim 1, The virtual link passing verification means that the proportion of sample objects in which the training results in the sample object set match the prediction results reaches a preset ratio threshold.

4. The method according to claim 1, Using the sample object set as a mock test case for the target organization to perform prediction based on the virtual link to test the prediction result of the target organization for the sample object set includes: Obtain feature data of the sample object set used by the target organization to train the local prediction model in the multi-party secure computing from the data management system of the target organization, and request the feature data of the used sample object set to the virtual link of the target organization to complete the prediction for the sample object set.

5. The method according to any one of claims 1 to 3, The training result corresponding to the sample object set is the training result corresponding to the last round of the multi-party secure computation performed by the target organization.

6. A method for configuring a multi-party prediction service in an organization, comprising: Before configuring an actual link for providing prediction services based on a local prediction model to a target organization, obtaining a set of sample objects used by the target organization to train the local prediction model in a multi-party secure computation and a training result corresponding to the set of sample objects; The sample objects of the sample object set are common sample objects determined by the target organization and other organizations based on the intersection of the privacy protection set, and the other organizations belong to the multi-party secure computation; Using the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model. Verifying the virtual link based on a degree of matching between a training result and a prediction result of the sample object set; When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

7. A verification device for a multi-party prediction service in an institution, comprising: A multi-party secure computing tracing module obtains a sample object set used by the target institution to train a local prediction model in a multi-party secure computing and the training results corresponding to the sample object set; The sample objects of the sample object set are common sample objects determined by the target organization and other organizations based on the intersection of the privacy protection set, and the other organizations belong to the multi-party secure computation; A mock testing module uses the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, so as to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model; The virtual link verification module verifies the virtual link based on the matching degree between the training result and the prediction result of the sample object set.

8. An electronic device comprising: A memory, a processor, and a computer program stored in the memory and operable on the processor, wherein the computer program is executed by the processor: Obtaining a sample object set used by the target organization for training a local prediction model in a multi-party secure computation and a training result corresponding to the sample object set; The sample objects of the sample object set are common sample objects determined by the target organization and other organizations based on the intersection of the privacy protection set, and the other organizations belong to the multi-party secure computation; Using the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model. The virtual link is verified based on a matching degree between a training result and a prediction result of the sample object set.

9. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the following steps: Obtaining a sample object set used by a target organization in a multi-party secure computation to train a local prediction model and a training result corresponding to the sample object set; the sample objects in the sample object set are shared sample objects determined by the target organization and other organizations based on the intersection of privacy-preserving sets, and the other organizations are members of the multi-party secure computation; Using the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization to provide prediction services after completing multi-party secure computation training on a local prediction model. The virtual link is verified based on a matching degree between a training result and a prediction result of the sample object set.

10. A device for configuring a multi-party prediction service in an organization, comprising: The multi-party secure computing tracing module obtains the sample object set used by the target organization to train the local prediction model in the multi-party secure computing and the training results corresponding to the sample object set before configuring the actual link for providing prediction services based on the local prediction model to the target organization; The sample objects of the sample object set are common sample objects determined by the target organization and other organizations based on the intersection of the privacy protection set, and the other organizations belong to the multi-party secure computation; A mock testing module uses the sample object set as a mock test case for the target organization to perform prediction based on a virtual link, so as to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization after completing multi-party secure computation training on a local prediction model to simulate the actual link after configuration to provide prediction services; A virtual link verification module verifies the virtual link based on a matching degree between a training result and a prediction result of the sample object set; The actual link configuration module configures an actual link for providing prediction services to the target organization according to the virtual link when the virtual link passes the verification.

11. An electronic device comprising: A memory, a processor, and a computer program stored in the memory and operable on the processor, wherein the computer program is executed by the processor: Before configuring an actual link for providing prediction services based on a local prediction model to a target organization, obtaining a set of sample objects used by the target organization to train the local prediction model in a multi-party secure computation and a training result corresponding to the set of sample objects; The sample objects of the sample object set are common sample objects determined by the target organization and other organizations based on the intersection of the privacy protection set, and the other organizations belong to the multi-party secure computation; Using the sample object set as a mock test case for the target organization's prediction based on a virtual link to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization after completing multi-party secure computation training on a local prediction model to simulate the actual link after configuration to provide prediction services; Verifying the virtual link based on a degree of matching between a training result and a prediction result of the sample object set; When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

12. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the following steps: Before configuring an actual link for providing prediction services based on a local prediction model to a target organization, obtaining a set of sample objects used by the target organization to train the local prediction model in a multi-party secure computation and a training result corresponding to the set of sample objects; The sample objects of the sample object set are common sample objects determined by the target organization and other organizations based on the intersection of the privacy protection set, and the other organizations belong to the multi-party secure computation; Using the sample object set as a mock test case for the target organization's prediction based on a virtual link to test the target organization's prediction results for the sample object set. The virtual link is a link constructed by the target organization after completing multi-party secure computation training on a local prediction model to simulate the actual link after configuration to provide prediction services; Verifying the virtual link based on a degree of matching between a training result and a prediction result of the sample object set; When the virtual link passes the verification, the actual link for providing the prediction service to the target organization is configured according to the virtual link.

Images