Calculation system, calculation method, and program

The calculation system enhances security in federated learning by distributing the management and calculation of secret information products using rearrangement and encryption, preventing leakage to third parties.

JP2024159627AActive Publication Date: 2024-11-08RAKUTEN GROUP INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2024070185
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-04-28
Filing Date
2024-04-24
Publication Date
2024-11-08
Estimated Expiration
2044-04-24

AI Technical Summary

Technical Problem

In federated learning and other scenarios where secret information from multiple devices needs to be combined, the security of transmitting and calculating the product of secret information is insufficient, as the information can be decoded by third parties even when encrypted.

Method used

A calculation system that includes devices managing secret information and sorting information, allowing for the distributed management and calculation of secret products without direct transmission, using rearrangement and encryption techniques to enhance security.

Benefits of technology

The system increases security by enabling the calculation of secret products without direct transmission, reducing the risk of information leakage to third parties, even when encrypted information is intercepted.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024159627000001_ABST
    Figure 2024159627000001_ABST
Patent Text Reader

Abstract

To enhance security.SOLUTION: Provided is a computing system including: a first device configured to secretly manage first secret information and first permutation information; and a second device configured to secretly manage second secret information and second permutation information. The first device acquires first managed information that is managed by the first device from among first permutation results managed in a dispersed manner by the first device and the second device, and acquires second managed information that is managed by the first device from among second permutation results managed in a dispersed manner by the first device and the second device. The second device acquires third managed information that is managed by the second device from among the first permutation results, and acquires fourth managed information that is managed by the second device from among the second permutation results. The computing system calculates a secret product which is a product of the first secret information and the second secret information on the basis of the first managed information, the second managed information, the third managed information and the fourth managed information.SELECTED DRAWING: Figure 5
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates to a computing system, a computing method, and a program. [Background technology]

[0002] Conventionally, techniques for preventing leakage of information secretly managed by a device have been considered. For example, Non-Patent Document 1 describes a federated learning method for creating a learning model using a machine learning method without consolidating training data at one location. In federated learning, it is known that each of a plurality of devices can execute learning of a learning model without transmitting information secretly managed by itself to other devices. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Qinbin Li, Zeyi Wen, Zhaomin Wu, Sixu Hu, Naibo Wang, Yuan Li, Xu Liu, Bingsheng He, “A Survey on Federated Learning Systems: Vision, Hype and Reality for Data Privacy and Protection”, Internet, retrieved March 16, 2023, https: / / arxiv.org / abs / 1907.09693 Summary of the Invention [Problem to be solved by the invention]

[0004] For example, in associative learning such as that described in non-patent literature, there is a demand for calculating a secret product of first secret information secretly managed by a first device and second secret information secretly managed by a second device in order to learn a learning model. In fields other than associative learning, there is a similar demand when a secret product, which is a product of multiple pieces of information secretly managed, is calculated. In this case, however, when the second device calculates the secret product of the first secret information and the second secret information, if the first device transmits the first secret information to the second device, the first secret information is transmitted over a network, and security is not sufficient. Even if the first secret information is encrypted, there is a possibility that it can be decrypted by a third party. The same applies when the first device calculates the secret product.

[0005] One goal of this disclosure is to increase security. [Means for solving the problem]

[0006] A computing system according to the present disclosure includes a first device that secretly manages first secret information of a number of elements that is 2 or more and first permutation information for permuting the elements of the number of elements, and a second device that secretly manages second secret information of the number of elements and second permutation information for permuting the elements of the number of elements, wherein the first device includes a first management information acquisition unit that acquires first management information managed by the first device from a first permutation result that is a result of permutation of the first secret information with the first permutation information and the second permutation information and is managed in a distributed manner by the first device and the second device, and a second management information acquisition unit that acquires first management information managed by the first device from the first permutation result that is a result of permutation of the first secret information with the first permutation information and the second permutation information, and and a second management information acquisition unit that acquires second management information managed by the first device from a second sorting result, the second sorting result being a result of sorting secret information and being managed in a distributed manner by the first device and the second device. The second device includes a third management information acquisition unit that acquires third management information managed by the second device from the first sorting result, and a fourth management information acquisition unit that acquires fourth management information managed by the second device from the second sorting result. A secret product that is a product of the first secret information and the second secret information is calculated based on the first management information, the second management information, the third management information, and the fourth management information. Effect of the Invention

[0007] The present disclosure provides increased security. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 illustrates an example of federated learning performed in a computing system. [Diagram 2] FIG. 13 is a diagram illustrating an example of a calculation formula in associative learning. [Diagram 3] FIG. 2 is a diagram showing a specific example of the problem in the table of FIG. 1. [Figure 4] FIG. 2 illustrates an example of a hardware configuration of a computing system. [Diagram 5] FIG. 2 is a diagram illustrating an example of functions realized in a computing system. [Figure 6] 13 is a diagram showing an example of rearrangement of first secret information executed based on first rearrangement information and second rearrangement information. FIG. [Figure 7] FIG. 11 is a diagram showing an example of how first secret information is shared. [Figure 8] FIG. 13 is a diagram showing an example of a technical idea for obtaining an inner product of first secret information and second secret information. [Figure 9] FIG. 2 illustrates an example of a process executed in a computing system. [Figure 10] FIG. 11 is a diagram showing an example of details of the processes in S20 and S21. [Figure 11] FIG. 11 is a diagram showing an example of the details of the processes in S200 to S203. [Figure 12] FIG. 11 is a diagram showing an example of details of the processes in S30 and S31. [Figure 13] FIG. 11 is a diagram showing an example of the details of the processes in S300 to S303. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] [1. Overview of the computation system] The present embodiment, which is an example of an embodiment of a computing system according to the present disclosure, will be described. In the present embodiment, a case where a computing system is used for federated learning in machine learning is taken as an example. Federated learning is a learning method that is executed in a distributed manner by multiple devices. Any one of the multiple devices aggregates learning results from the other devices to create a learning model.

[0010] In conventional learning, a device that performs learning of a learning model needs to collect training data from other devices. This raises security concerns because there is a possibility that the training data may be leaked to a third party. In this regard, in federated learning, each device does not, in principle, send training data to other devices, but only the learning results, which is useful from a security perspective. For example, if the training data includes personal information, federated learning can prevent the leakage of personal information.

[0011] FIG. 1 is a diagram showing an example of federated learning executed in a computing system. In the example of FIG. 1, a case is shown in which four services 0 to 3 cooperate with each other to execute federated learning. For example, service 0 is an information management service provided by a management company that manages various information related to users. service 1 is a financial service provided by a bank. service 2 is a financial service provided by a securities company. service 3 is an insurance service provided by an insurance company.

[0012] For example, computers in each of services 0 to 3 cooperate with each other to perform federated learning. In Fig. 1, the computers are omitted. Hereinafter, for the sake of simplicity in Fig. 1, the explanation will be given assuming that the executing entities that execute the processing in the federated learning are services 0 to 3, but in reality, the executing entities are the computers in each of services 0 to 3. Hereinafter, when there is no need to distinguish between services 0 to 3, the reference numerals will be omitted and they will be referred to as services.

[0013] For example, in federated learning, each service initializes its own learning model. Each service executes learning of its own learning model based on training data managed by itself. In principle, the training data is not disclosed to third parties. The training data is managed confidentially by each service. In this embodiment, an example is given in which a learning model that estimates a user's annual income is created by federated learning, but a learning model may be created for any other purpose. For example, a learning model that estimates attributes other than annual income, a learning model used in image processing, a learning model used in natural language processing, or other learning models may be created by federated learning.

[0014] For example, the training data includes demographic information that is considered to be correlated with annual income, payment history in the service, web page browsing history, or other information. Each service manages the training data independently, and does not transmit the training data to the outside in principle. In this embodiment, a case where the training data shows the characteristics of a user who uses all of services 1 to 3 is taken as an example, but it is sufficient that the training data shows the characteristics of a user who uses at least one of services 1 to 3.

[0015] FIG. 2 is a diagram showing an example of a formula in federated learning. For example, service 1 executes learning of a learning model M1 that estimates a user's annual income based on training data indicating user characteristics x1 that are secretly managed by a bank. Service 2 executes learning of a learning model M2 that estimates a user's annual income based on training data indicating user characteristics x2 that are secretly managed by a securities company. Service 3 executes learning of a learning model M3 that estimates a user's annual income based on training data indicating user characteristics x3 that are secretly managed by an insurance company.

[0016] In the example of Figure 2, a concept that includes features x1, x2, and x3 is described as feature X. Feature X is in the form of an n x m matrix, where n and m are any natural numbers. For example, service 0 manages label y, which is the correct answer for annual income. In the example of Figure 2, y is an n x 1 matrix (a so-called column vector). Service 0 collects the local learning results from each of services 1 to 3, and executes learning of a learning model M0 that estimates the user's annual income.

[0017] In the example of FIG. 1, the service 0 creates a learning model M0 that estimates the user's annual income y based on the estimation result y1 by the learning model M1, the estimation result y2 by the learning model M2, and the estimation result y3 by the learning model M3. That is, the estimation results y1 to y3 are input to the learning model M0. The learning model M0 calculates features based on the estimation results y1 to y3, and outputs the user's annual income y as an estimation result according to the features. The learning model M0 can be said to be a model in stacking learning, which is a type of ensemble learning that uses the estimation results y1 to y3 of the learning models M1 to M3.

[0018] For example, the learning model M0 includes weighting coefficients w0 to w4 as parameters adjusted by learning. The learning model M0 may include parameters other than the weighting coefficients w0 to w4 (e.g., bias). In FIG. 2, i is a numerical value indicating an individual user. As shown in FIG. 2, the correct label y i and the estimation result y of the learning model M0 i (In Figure 2, a symbol is added at the top) and the loss L is calculated based on the difference between them. Service 0 performs training of the learning model M0 so that the loss L becomes small.

[0019] For example, service 0 transmits the learning results of learning model M0 to each of services 1 to 3. Service 1 updates learning model M1 based on the learning results of learning model M0. Service 2 updates learning model M2 based on the learning results of learning model M0. Service 3 updates learning model M3 based on the learning results of learning model M0. Each of services 1 to 3 transmits new learning results to service 0. In the same manner, the sharing of learning results between service 0 and each of services 1 to 3 is repeated, and improvement of learning models M0 to M3 is expected.

[0020] In FIG. 1, lines are omitted for simplicity, but in this embodiment, services 1 to 3 share each other's learning results. For example, service 1 obtains the learning results of learning models M2 and M3 from services 2 and 3, and updates learning model M1. Service 2 obtains the learning results of learning models M1 and M3 from services 1 and 3, and updates learning model M2. Service 3 obtains the learning results of learning models M1 and M2 from services 1 and 2, and updates learning model M3.

[0021] Hereinafter, when the learning models M0 to M4 are not distinguished from each other, they are simply referred to as the learning model M. The learning method of the learning model M may be a known method, and for example, the gradient descent method or the backpropagation method may be used. In the present embodiment, the type and the learning method of the learning models M are the same as each other, but at least one of the type and the learning method of the learning models M may be different from each other. For example, the learning model M0 may be a neural network, the learning model M1 may be a support vector machine, the learning model M2 may be a gradient boosting regression tree, and the learning model M3 may be a linear regression model. The learning method may be a method according to these types.

[0022] In this embodiment, the learning model M is a neural network, and the learning method of the learning model M is the gradient descent method. The gradient descent method iteratively updates the parameters of the learning model M, calculates the gradient (slope) of the loss, and updates the parameters in the direction in which the gradient falls the most. For example, when the gradient descent method is used in the learning of the learning model M0, a differential calculation formula as shown in FIG. 2 is used. When the gradient descent method is used in the learning of the learning models M1 to M3, a similar calculation formula is used. Since the gradient descent method itself is a known technology, details of the calculation formula in FIG. 2 are omitted.

[0023] For example, the underlined part in the formula in Figure 2 is the inner product of the column vector related to service 1 and the column vector related to service 2. When services 1 and 2 execute learning of learning models M1 and M2, respectively, they need to obtain similar information. In this case, services 1 and 2 are required to obtain the inner product underlined in the formula in Figure 2 without transmitting information that should be kept secret between each other. Table C1 in Figure 1 generalizes the problem of satisfying such a requirement. If a computing system can solve the problem in Table C1, it can obtain the inner product without transmitting information that should be kept secret between the services.

[0024] FIG. 3 is a diagram showing a specific example of the problem of Table C1 in FIG. 1. For example, a first device A of a certain service secretly manages an m-dimensional vector a. A second device B of another service secretly manages an m-dimensional vector b. For example, an element included in the m-dimensional vector indicates some characteristic of an individual user. The second device B is required to have high security. From the viewpoint of security, each of the first device A and the second device B has a constraint that some integrated information is transmitted, instead of the vector a and the vector b.

[0025] For example, as shown in Table C2, if the second device B can obtain the inner product a·b of vector a and vector b without obtaining vector a from the first device A, this is useful for learning the learning model M managed by the second device B. In this embodiment, an example of this method is described. Conversely, as shown in Table C3, if the first device A can obtain the inner product a·b of vector a and vector b without obtaining vector b from the second device B, this is useful for learning the learning model M managed by the first device A. The computation system 1 is also capable of such a method.

[0026] [2. Hardware configuration of the computing system] 4 is a diagram illustrating an example of a hardware configuration of a computing system. A first device A and a second device B are each connected to a network N such as the Internet or a LAN.

[0027] The first device A is a computer for the first service. In this embodiment, the first service is taken as an example to be service 1 in FIG. 1 (a service provided by a bank), but the first service may be any other service. For example, the first service may be service 0, 2, or 3 in FIG. 1, or may be another service not shown in FIG. 1. The other service may be a communication service, an electronic commerce service, a travel reservation service, an electronic payment service, or a point management service.

[0028] For example, the first device A is a server computer, a personal computer, a tablet, or a smartphone. The first device A includes a control unit A11, a storage unit A12, and a communication unit A13. The control unit A11 includes at least one processor. The storage unit A12 includes a volatile memory such as a RAM, and a non-volatile memory such as a flash memory. The communication unit A13 includes at least one of a communication interface for wired communication and a communication interface for wireless communication.

[0029] The second device B is a computer of the second service. For example, the second device B is a server computer, a personal computer, a tablet, or a smartphone. The second device B includes a control unit B11, a storage unit B12, and a communication unit B13. The hardware configurations of the control unit B11, the storage unit B12, and the communication unit B13 may be similar to those of the control unit A11, the storage unit A12, and the communication unit A13, respectively.

[0030] In this embodiment, the second service is taken as an example to be service 2 in FIG. 1 (a service provided by a securities company), but the second service may be any other service. For example, the second service may be the same as the first service. When the second service is different from the first service, the second service may be a service other than the first service among the services exemplified in the description of the first service. The first device A and the second device B may be devices used for private or other purposes, regardless of any service.

[0031] In this embodiment, a case is taken as an example where higher security is required for the second device B than for the first device A. For example, limiting the data that can be transmitted to the network N corresponds to high security. When the second device B transmits the same data as the first device A to the outside, the second device B needs to process the data more than the first device A. For example, the first device A can transmit data in a matrix format to the network N without changing the number of rows, but the second device B needs to change the number of rows of the data in a matrix format before transmitting it to the network N.

[0032] Hereinafter, when the first device A and the second device B are not distinguished from each other, they are simply referred to as devices and the reference numerals are omitted. When the control units A11 and B11, the storage units A12 and B12, and the communication units A13 and B13 are not distinguished from each other, they are respectively referred to as control unit 11, storage unit 12, and communication unit 13 and the alphabet at the beginning of the reference numerals is omitted. The program stored in the storage unit 12 may be supplied to the device via the network N. In addition, the program stored in a computer-readable information storage medium may be supplied to the device via a reading unit (e.g., an optical disk drive or a memory card slot) that reads the information storage medium, or an input / output unit (e.g., a USB port) that inputs and outputs data to and from an external device.

[0033] In addition, in the computing system 10, each of a plurality of devices including the first device A, the second device B, and other devices may be capable of communication. That is, the computing system 10 may include other devices other than the first device A and the second device B. For example, the other devices are a server computer, a personal computer, a tablet, or a smartphone. The computing system 10 may include a plurality of other devices. When there are four services as in FIG. 1, the computing system 10 includes the first device A, the second device B, and at least two other devices.

[0034] For example, when the computing system 10 includes other devices, the second device B is required to have relatively high security among the multiple devices. The other devices may have a function similar to the first device A or the second device B. When the other devices have a function similar to the first device A, the inclusion of the first device A and the other devices in the computing system 10 means that the computing system 10 includes a plurality of first devices A. When the other devices have a function similar to the second device B, the inclusion of the second device B and the other devices in the computing system 10 means that the computing system 10 includes a plurality of second devices B.

[0035] [3. Functions realized by the computational system] 5 is a diagram showing an example of functions realized by the computing system 10. For example, the first device A includes a first data storage unit A100, a first secret information acquisition unit A101, a first rearrangement information acquisition unit A102, a first management information acquisition unit A103, a second management information acquisition unit A104, a fifth management information acquisition unit A105, a first public information reception unit A106, a tenth management information acquisition unit A107, a second public information acquisition unit A108, a second public information transmission unit A109, a fifth management information transmission unit A110, a first key acquisition unit A111, a first public key transmission unit A112, an eleventh management information acquisition unit A113, a twelfth management information transmission unit A114, a tenth management information transmission unit A115, a first public key transmission unit A116, a second public key transmission unit A117, a third management information transmission unit A118, a fourth management information transmission unit A119, a fifth management information transmission unit A120, a fifth management information transmission unit A210, a fifth management information transmission unit A211, a fifth management information transmission unit A212, a fifth management information transmission unit A213, a fifth management information transmission unit A214, a fifth management information transmission unit A215, a fifth management information transmission unit A216, a fifth management information transmission unit A217, a fifth management information transmission unit A218, a fifth management information transmission unit A219, a fifth management information transmission unit A220, a fifth management information transmission unit A221, a fifth management information transmission unit A222, a fifth management information transmission unit A223, a fifth management information transmission unit A224, a fifth management information transmission unit A225, a fifth management information transmission unit A226, a fifth management information transmission unit A227, a fifth management information transmission unit A228, a fifth management information transmission unit A229, a fifth management information transmission unit A230, a fifth management information transmission unit A231, a fifth The cryptographic unit A120 includes a management information acquisition unit A114, a third public information receiving unit A115, a second calculation information acquisition unit A116, a sixteenth management information acquisition unit A117, a fourth public information acquisition unit A118, a fourth public information transmission unit A119, a thirteenth management information receiving unit A120, a sixth sorting result acquisition unit A121, a fifteenth management information acquisition unit A122, a fifteenth management information transmission unit A123, a second public key receiving unit A124, a first management product acquisition unit A125, a first total acquisition unit A126, a second total receiving unit A127, and a secret product calculation unit A128.

[0036] For example, the first data storage unit A100 is realized by the storage unit A12. A first secret information acquisition unit A101, a first rearrangement information acquisition unit A102, a first management information acquisition unit A103, a second management information acquisition unit A104, a fifth management information acquisition unit A105, a first public information reception unit A106, a tenth management information acquisition unit A107, a second public information acquisition unit A108, a second public information transmission unit A109, a fifth management information transmission unit A110, a first key acquisition unit A111, a first public key transmission unit A112, an eleventh management information acquisition unit A113, a twelfth management information acquisition unit A114, a third public information reception unit A115, a fourth management information acquisition unit A116, a fifth management information acquisition unit A117, a fifth management information acquisition unit A118, a fifth management information acquisition unit A119, a fifth management information acquisition unit A120, a fifth management information acquisition unit A210, a fifth management information acquisition unit A211, a fifth management information acquisition unit A212, a fifth management information acquisition unit A213, a fifth management information acquisition unit A214, a fifth management information acquisition unit A215, a sixth management information acquisition unit A216, a fifth management information acquisition unit A217, a sixth management information acquisition unit A218, a seventh management information acquisition unit A219, a eighth management information acquisition unit A220, a eighth management information acquisition unit A221, a seventh management information acquisition unit A222, a eighth management information acquisition unit A223, a eighth management information acquisition unit A224, a eighth management information acquisition unit A225, a eighth management information acquisition unit A226, a eighth management information acquisition unit A227, a eighth management information acquisition unit A228, a eighth management information acquisition unit A229, a eighth management information acquisition unit A230, a eighth management information acquisition unit A231, a eighth management information The control unit A11 realizes the second calculation information acquisition unit A115, the second calculation information acquisition unit A116, the sixteenth management information acquisition unit A117, the fourth public information acquisition unit A118, the fourth public information transmission unit A119, the thirteenth management information reception unit A120, the sixth rearrangement result acquisition unit A121, the fifteenth management information acquisition unit A122, the fifteenth management information transmission unit A123, the second public key reception unit A124, the first management product acquisition unit A125, the first total acquisition unit A126, the second total reception unit A127, and the secret product calculation unit A128.

[0037] For example, the second device B includes a second data storage unit B100, a second secret information acquisition unit B101, a second rearrangement information acquisition unit B102, a third management information acquisition unit B103, a fourth management information acquisition unit B104, a seventh management information acquisition unit B105, an eighth management information acquisition unit B106, a first public information acquisition unit B107, a first public information transmission unit B108, a second public information reception unit B109, a first calculation information acquisition unit B110, a fifth management information reception unit B111, a third rearrangement result acquisition unit B112, The communication device includes a sixth management information acquisition unit B113, a sixth management information transmission unit B114, a first public key receiving unit B115, a thirteenth management information acquisition unit B116, a fourteenth management information acquisition unit B117, a third public information acquisition unit B118, a third public information transmission unit B119, a fourth public information receiving unit B120, a third calculation information acquisition unit B121, a second key acquisition unit B122, a second public key transmission unit B123, a second management product acquisition unit B124, a second total acquisition unit B125, and a second total transmission unit B126.

[0038] For example, the second data storage unit B100 is realized by the storage unit B12. A second secret information acquisition unit B101, a second rearrangement information acquisition unit B102, a third management information acquisition unit B103, a fourth management information acquisition unit B104, a seventh management information acquisition unit B105, an eighth management information acquisition unit B106, a first public information acquisition unit B107, a first public information transmission unit B108, a second public information reception unit B109, a first calculation information acquisition unit B110, a fifth management information reception unit B111, a third rearrangement result acquisition unit B112, a sixth management information acquisition unit B113, a sixth management information transmission unit B114, a sixth management information acquisition unit B115, a sixth management information transmission unit B116, a seventh management information acquisition unit B117, a seventh management information acquisition unit B118, a seventh management information acquisition unit B119, a seventh management information acquisition unit B200, a seventh management information acquisition unit B201, a seventh management information acquisition unit B202, a seventh management information acquisition unit B203, a seventh management information acquisition unit B204, a seventh management information acquisition unit B205, a seventh management information acquisition unit B206, a seventh public information acquisition unit B207, a seventh public information acquisition unit B208, a seventh public information acquisition unit B209, a seventh calculation information acquisition unit B110, a seventh management information acquisition unit B209, a seventh management information acquisition unit B210, a seventh management information acquisition unit B211, a seventh management information acquisition unit B212, a seventh management information acquisition unit B213, a seventh management information acquisition unit B214, a seventh management information acquisition unit B215, a seventh management information acquisition unit B216, a seventh management information acquisition unit B217, a seventh management information acquisition unit B218, a seventh management information acquisition unit B219, a seventh management information acquisition unit 114, the first public key receiving unit B115, the 13th management information acquisition unit B116, the 13th management information transmitting unit, the 14th management information acquisition unit B117, the third public information acquisition unit B118, the third public information transmitting unit B119, the fourth public information receiving unit B120, the third calculation information acquisition unit B121, the second key acquisition unit B122, the second public key transmitting unit B123, the second management product acquisition unit B124, the second total acquisition unit B125, and the second total transmitting unit B126 are realized by the control unit B11.

[0039] [First data storage unit] The first data storage unit A100 stores data necessary for processing executed by the first device A. For example, the first data storage unit A100 stores a first learning model AM. The first learning model AM is a learning model M managed by the first device A. The first device A executes learning of the first learning model AM using an associative learning technique.

[0040] For example, the first device A secretly manages first secret information a having an element number i (i is an integer equal to or greater than 2). The first data storage unit A100 stores the first secret information a. The first secret information a is information secretly managed by the first device A. In principle, the first secret information a is not transmitted to devices other than the first device A. The first secret information a includes multiple elements. The elements are numerical values ​​constituting the first secret information a. In the present embodiment, a case where the first secret information a is in vector format is taken as an example, but the first secret information a may be any information that includes multiple elements and whose product can be calculated, and may be, for example, multiple numerical values, a matrix, an array, a data frame, or other information.

[0041] In this embodiment, the first device A calculates the first secret information a in the learning process of the first learning model AM in the associative learning. For example, the first device A may acquire a part of the parameters of the first learning model AM as the first secret information a, or may acquire information calculated from the part as the first secret information a. For example, the first secret information a is a value obtained by multiplying a part of the parameters of the current first learning model AM by the estimation result of the current first learning model AM.

[0042] In addition, the first secret information a may be information about each of a plurality of users in the first service. For example, the first secret information a may include a value obtained by multiplying a part of the parameters of the current first learning model AM by an estimated result of the annual income of each of the plurality of users by the current first learning model AM. In this case, the number of dimensions of the first secret information a is the number of users. For another example, the first secret information a may be information in which personal information such as the user's age group or gender is quantified (quantified). The first secret information a may be information in which each of a plurality of personal information of the user is quantified.

[0043] In addition, the user identification information of each of the multiple users in the first service and the user identification information of each of the multiple users in the second service may be the same. The user identification information is information that can identify a user. For example, a login account, an email address, or a phone number corresponds to the user identification information. In this embodiment, a case where the user identification information is common to the first service and the second service is taken as an example, but the user identification information may be different between the first service and the second service. In this case, data is assumed to exist that defines the correspondence between the user identification information in the first service and the user identification information in the second service for each user. This data is assumed to be stored in the first data storage unit A100, the second data storage unit B100, or other locations.

[0044] For example, the first device A generates first rearrangement information π A The first data storage unit A100 stores the first rearrangement information π A The first rearrangement information π A is information that is secretly managed by the first device A. A In principle, the first rearrangement information π is not transmitted to any device other than the first device A. A indicates the rule for rearrangement of elements. For example, the first rearrangement information π A indicates the relationship between the order of elements before rearrangement and the order of elements after rearrangement. In this embodiment, the first rearrangement information π A As an example, the first rearrangement information π A may indicate a rule for sorting elements, and may be, for example, a single numerical value, multiple numerical values, a matrix, a vector, a data frame, or other information.

[0045] The data stored in the first data storage unit A100 is not limited to the above example. The first data storage unit A100 can store any data. For example, the first data storage unit A100 stores training data of the first learning model AM. The training data includes an input portion input to the first learning model AM and an output portion corresponding to the input portion. The output portion corresponds to a correct answer during learning. The first device A performs local learning of the first learning model AM based on the training data stored in the first data storage unit A100. The first device A adjusts the parameters of the first learning model AM so that the output portion of the training data is output when the input portion of the training data is input.

[0046] In this embodiment, since a first learning model AM that estimates the user's annual income is created, the input part of the training data is some information about the user. For example, the characteristics of the user's demographic information in the first service correspond to the input part of the training data. The output part of the training data is the user's annual income. The annual income corresponding to the output part of the training data may be the annual income registered by the user in the first service, or the annual income registered by the user in another service. The annual income corresponding to the output part of the training data may be the annual income specified by the creator of the first learning model AM. The first device A may update the first learning model AM that has been locally learned using a federated learning technique. For example, the first data storage unit A100 may store various information described in this embodiment in addition to the data described above.

[0047] [Second data storage unit] The second data storage unit B100 stores data necessary for processing executed by the second device B. For example, the second data storage unit B100 stores a second learning model BM. The second learning model BM is a learning model M managed by the second device B. The second device B executes learning of the second learning model BM using an associative learning technique.

[0048] For example, the second device B secretly manages second secret information b having the number of elements i. The second data storage unit B100 stores the second secret information b. The second secret information b is information secretly managed by the second device B. In principle, the second secret information b is not transmitted to devices other than the second device B. The second secret information b is information indicating one numerical value or information including multiple numerical values. In this embodiment, a case where the second secret information b is in a vector format is taken as an example, but the second secret information b may be any information that includes multiple elements and whose product can be calculated, and may be, for example, multiple numerical values, a matrix, an array, a data frame, or other information.

[0049] In this embodiment, the second device B calculates the second secret information b in the learning process of the second learning model BM in the associative learning. For example, the second device B may acquire a part of the parameters of the second learning model BM as the second secret information b, or may acquire information calculated from the part as the second secret information b. For example, the second secret information b is a value obtained by multiplying a part of the parameters of the current second learning model BM by the estimation result of the current second learning model BM.

[0050] In addition, the second secret information b may be information about each of the multiple users in the second service. For example, the second secret information b may include a value obtained by multiplying a part of the parameters of the current second learning model BM by the estimated annual income of each of the multiple users by the current second learning model BM. In this case, the number of dimensions of the second secret information b is the number of users. For another example, the second secret information b may be information in which personal information such as the user's age group or gender is quantified. The second secret information b may be information in which each of the user's multiple personal information is quantified.

[0051] For example, the second device B generates second rearrangement information π B The second data storage unit B100 stores the second rearrangement information π B The second rearrangement information π B is information that is secretly managed by the second device B. The second permutation information π BIn principle, the second rearrangement information π is not transmitted to any device other than the second device B. B indicates the rule for rearrangement of elements. For example, the second rearrangement information π B indicates the relationship between the order of elements before rearrangement and the order of elements after rearrangement. In this embodiment, the second rearrangement information π B As an example, let us consider a case where the second sorting information π B may indicate a rule for sorting elements, and may be, for example, a single numerical value, multiple numerical values, a matrix, a vector, a data frame, or other information.

[0052] The data stored in the second data storage unit B100 is not limited to the above example. The second data storage unit B100 can store any data. For example, the second data storage unit B100 stores training data of the second learning model BM. The training data includes an input portion input to the second learning model BM and an output portion corresponding to the input portion. The output portion corresponds to the correct answer during learning. The second device B performs local learning of the second learning model BM based on the training data stored in the second data storage unit B100. The second device B adjusts the parameters of the second learning model BM so that the output portion of the training data is output when the input portion of the training data is input.

[0053] In this embodiment, since the second learning model BM that estimates the user's annual income is created, the input part of the training data is some information about the user. For example, the characteristics of the user's demographic information in the second service correspond to the input part of the training data. The output part of the training data is the user's annual income. The annual income corresponding to the output part of the training data may be the annual income registered by the user in the second service, or the annual income registered by the user in another service. The annual income corresponding to the output part of the training data may be the annual income specified by the creator of the second learning model BM. The second device B may update the second learning model BM learned locally using a federated learning technique. For example, the second data storage unit B100 may store various information described in this embodiment in addition to the data described above.

[0054] Hereinafter, with reference to FIGS. 6 to 8, a technical concept regarding the processing executed by functions other than the first data storage unit A100 and the second data storage unit B100 will be described.

[0055] FIG. 6 shows the first rearrangement information π A and second sorting information π B 6 is a diagram showing an example of rearrangement of the first secret information a executed based on the above. The rearrangement in FIG. 6 is not a process actually executed by the first device A and the second device B. The rearrangement in FIG. 6 is an example for explaining the above technical idea. In the example in FIG. 6, the first secret information a is (5, 30, 65, 6, 62). The number of elements i is 5. The first rearrangement information π A is (1,3,5,2,4). An array may be expressed in square brackets, but in the example of FIG. 6, it is expressed in small brackets. The second permutation information π B is (2,5,1,3,4).

[0056] For example, the first rearrangement information π A and second sorting information π B Each element of the first sorting information π indicates the order after sorting. A The m-th element (m is a natural number equal to or smaller than i) of the first permutation information π is the order after the m-th element of the first secret information a is permuted. A Since is (1,3,5,2,4), the first element of the first secret information a will become the first after rearrangement. The second element of the first secret information a will become the third after rearrangement. The third element of the first secret information a will become the fifth after rearrangement. The fourth element of the first secret information a will become the fourth after rearrangement.

[0057] In this embodiment, the first rearrangement information π A The result of rearranging the first secret information a based on A (a). Other sortings are also indicated by symbols according to the same rule. The sorting result is the state after the information to be sorted is sorted. In the example of FIG. 6, the first sorting information π A The first secret information a after being rearranged based onA Corresponding to (a). The rearrangement result π A (a) becomes (5, 6, 30, 62, 65).

[0058] In this embodiment, the second rearrangement information π B The sorting rule based on A The first rearrangement information π A The sorting rule based on and the second sorting information π B In the example of FIG. 6, the rule of rearrangement based on the second rearrangement information π B Sorting results based on π A (a) is the rearranged result π B (π A (a)) becomes (30, 5, 62, 65, 6). If the sorting information indicating the final sorting is π, then π is (2, 1, 4, 5, 3). Note that π in FIG. A π B The symbol · in the above is not a product but is used to indicate the final sort order.

[0059] 7 is a diagram showing an example of how the first secret information a is shared. In this embodiment, the first secret information a, the second secret information b, and the first rearrangement information π A , and second sorting information π B A secret sharing scheme is used to calculate the inner product of the first secret information a and the second secret information b without transmitting each of them to the network N. The secret sharing scheme is a method of sharing fragments of random numbers that have no meaning in themselves. When all or part of the shares are collected, the original data is restored. In this embodiment, the technical idea of ​​rearrangement is applied to the well-known secret sharing scheme, and therefore the present embodiment has an innovative configuration compared to conventional techniques.

[0060] In the example of FIG. 7, the first secret information a is (3, 2, 2, 1, 1). IDs 1 to 5 written in the first secret information a are user identification information. For example, the first element of the first secret information a is information about the user of ID 1. Similarly, the second to fifth elements of the first secret information a are information about the users of ID 2 to ID 5, respectively. IDs 1 to 5 written in the second secret information b in FIG. 8 described later have the same meaning as IDs 1 to 5 written in the first secret information a. Therefore, the user corresponding to the m-th element of the first secret information a and the user corresponding to the mb-th element of the second secret information b are the same user.

[0061] In the example of FIG. 7, the first rearrangement information π A is (1,3,5,2,4). The second rearrangement information π B As shown in FIG. 7, if the first secret information a is shared by the secret sharing scheme, the first device A receives the information A The second device B manages the information B In this manner, in this embodiment, information shared by the secret sharing scheme is enclosed in the symbols <>, and the entity managing the information is indicated by the letter A or B after the symbols. Information shared by the secret sharing scheme is sometimes simply called a share. Information A and information B Although each of these is a fragment of a random number, when these are collected, it is possible to restore the first secret information a. The secret sharing scheme itself can use a publicly known method.

[0062] For example, information managed by the first device A A is (5,9,6,2,4). The first rearrangement information π A Information A The sorted result is <π A (a)> A is (5,2,9,4,6). If the second permutation information π B Information at <π A (a)> A If is rearranged, the rearrangement result < π B <π A (a))> A The first device A receives the second permutation information π B Since the first device A cannot obtain the data, this rearrangement is not directly performed by the first device A. The rearrangement will be described merely for the purpose of explaining the technical concept of this embodiment.

[0063] For example, information managed by the second device A B is (8,3,6,9,7). If the first permutation information π A Information B If is rearranged, the rearrangement result < π A (a)> B The second device B receives the first rearrangement information π A Since the second device B cannot obtain the second rearrangement information π, the rearrangement is not directly performed by the second device B. The rearrangement is described only for the purpose of explaining the technical concept of the present embodiment. B Information at <π A (a)> B The sorted result is <π B <π A (a))> B is (3,8,7,6,9).

[0064] 7, the first secret information a is shared between the first device A and the second device B after the technical idea of ​​rearrangement is applied. Similarly, the second secret information b is shared between the first device A and the second device B after the technical idea of ​​rearrangement is applied. If these shares are successful, at least one of the first device A and the second device B can obtain the inner product of the first secret information a and the second secret information b. In this embodiment, the case where the first device A obtains the inner product of the first secret information a and the second secret information b is taken as an example, but the second device B may obtain the inner product of the first secret information a and the second secret information b.

[0065] FIG. 8 is a diagram showing an example of a technical idea for obtaining the inner product of the first secret information a and the second secret information b. In the example of FIG. 8, the second secret information b is (1, 3, 1, 2, 3). The inner product of the first secret information a and the second secret information b is 16. For example, if the remainder when dividing by 10 is used in learning, the final inner product is 6. As shown in FIG. 8, in this embodiment, the problem of calculating the inner product of the first secret information a and the second secret information b is solved by the first device A by rearranging the result <π. B <π A (a))> A ,<π B <π A (b))> A and the second device B obtains the sorting result < π B <π A (a))> B ,<π B <π A (b))> B This is defined as the problem of obtaining

[0066] Hereinafter, an example of a process for solving this problem will be described, that is, an example of a process executed by a function other than the first data storage unit A100 and the second data storage unit B100 will be described.

[0067] FIG. 9 is a diagram showing an example of processing executed in the computing system 10. The processing in FIG. 9 corresponds to a main routine. The processing in FIG. 10 and subsequent figures correspond to subroutines. That is, the processing in FIG. 9 is an overall processing executed in the computing system 10. The processing in FIG. 10 and subsequent figures are specific examples of individual processing. The processing in FIG. 9 is executed by the first device A executing a program stored in the first data storage unit A100, and the second device B executing a program stored in the second data storage unit B100.

[0068] For example, the first secret information acquisition unit A101 acquires the first secret information a stored in the first data storage unit A100 (S10). By the process of S10, the first secret information a is read and expanded in a storage area of ​​a non-volatile memory. If the first secret information a has been expanded in the storage area in advance, the process of S10 does not need to be executed.

[0069] For example, the first rearrangement information acquisition unit A102 acquires the first rearrangement information π A In S11, the first rearrangement information acquisition unit A102 acquires the first rearrangement information π A For example, the first permutation information acquisition unit A102 generates random numbers the same number as the number of elements i, and generates the first permutation information π A The elements of the random numbers are as follows. A publicly known algorithm can be used as the algorithm for generating random numbers. For example, random numbers may be generated based on the linear congruential method, the multiplicative congruential method, the M-sequence method, or the Mersenne Twister method. When random numbers are required in subsequent processing, the first device A can generate random numbers based on these methods.

[0070] The first rearrangement information acquisition unit A102 does not use a random number but instead obtains the first rearrangement information π A For example, the first rearrangement information acquisition unit A102 may acquire the first rearrangement information π A Alternatively, only some elements of the first rearrangement information π may be obtained using random numbers, and the other elements may be set to predetermined fixed values. A All elements of the first permutation information π may be obtained without being based on random numbers. A may be stored in advance in the first data storage unit A100 instead of being generated on the spot. In this case, the first rearrangement information acquisition unit A102 refers to the first data storage unit A100 and obtains the first rearrangement information π A You just need to obtain the following.

[0071] For example, the second secret information acquisition unit B101 acquires the second secret information b stored in the second data storage unit B100 (S12). By the process of S12, the second secret information b is read and expanded in a storage area of ​​a non-volatile memory. If the second secret information b has been expanded in the storage area in advance, the process of S12 does not need to be executed.

[0072] For example, the second rearrangement information acquisition unit B102 acquires the second rearrangement information π B In S13, the second rearrangement information acquisition unit B102 acquires the second rearrangement information π B For example, the second rearrangement information acquisition unit B102 generates random numbers the same number as the number of elements i, and generates second rearrangement information π B The elements of the random numbers are as follows. A publicly known algorithm can be used as the algorithm for generating random numbers. For example, random numbers may be generated based on the linear congruential method, the multiplicative congruential method, the M-sequence method, or the Mersenne Twister method. When random numbers are required in subsequent processing, the second device B can generate random numbers based on these methods.

[0073] The second rearrangement information acquisition unit B102 does not use a random number but instead obtains the second rearrangement information π B For example, the second rearrangement information acquisition unit B102 may acquire the second rearrangement information π B Alternatively, only some elements of the second rearrangement information π may be obtained using random numbers, and the other elements may be set to predetermined fixed values. B All elements of the second permutation information π may be obtained without being based on random numbers. B may be stored in advance in the second data storage unit B100 instead of being generated on the spot. In this case, the second rearrangement information acquisition unit B102 refers to the second data storage unit B100 and obtains the second rearrangement information π B You just need to obtain the following.

[0074] For example, the first management information acquisition unit A103 acquires the first rearrangement information π A and second sorting information π B The first secret information a is rearranged, and the first rearrangement result (π B (π A Among (a), the first management information managed by the first device A <c> A =<π B (π A (a))> A In this embodiment, the first rearrangement result (π B (π A (a)) and the first management information <c> A Let us explain the case where is in vector form. The first rearrangement result (π B (π A (a)) and the first management information <c> A may be a number of numbers, a matrix, an array, a vector, a data frame, or other information.

[0075] For example, the third management information acquisition unit B103 acquires the first rearrangement result π B (π A (a)) the third management information managed by the second device B <c> B =<π B (π A (a))> B In the processes of S20 and S21, the first device A and the second device B each obtain the first rearrangement information π A and second sorting information π B This is a process for sharing the first secret information a using a secret sharing scheme by using rearrangement of

[0076] 10 is a diagram showing an example of the details of the processes of S20 and S21. For example, the fifth management information acquisition unit A105 acquires the fifth management information S managed by the first device A from the first distributed information S managed in a distributed manner by the first device A and the second device B. <s> A (S200). The first shared information S is information to be shared in the secret sharing scheme. In this embodiment, the first shared information S is a random number, but the first shared information S may be any information and is not limited to a random number. For example, the first shared information S may be a predetermined fixed value or the current time.

[0077] A detailed example of the process of S200 will be described later with reference to Fig. 11. Similarly, a detailed example of processes of S201 to S203 will be described later with reference to Fig. 11. Fifth Management Information <s> A The process for acquiring the fifth management information is not limited to the example shown in FIG. 11. The fifth management information acquiring unit A105 may acquire the fifth management information based on a secret sharing scheme other than that shown in FIG. <s> A The secret sharing scheme itself may be a known method, such as Shamir's Secret Sharing or Blakley's Secret Sharing. This also applies to the cases where any information is shared in the following description.

[0078] For example, the first management information acquisition unit A103 acquires the first shared information S from the second rearrangement information π B The third sorted result π is managed in a distributed manner by the first device A and the second device B. B Among (S), the sixth management information managed by the first device A<π B (S)> A , the first management information <π B (π A (a))> A The first device A and the second device B actually obtain the third sorting result π B (S) is not obtained, but the third sorting result π B For example, the seventh management information acquisition unit B105 acquires only the share of the seventh management information (S) managed by the second device B from the first distributed information S. <s> B (S202). The eighth management information acquisition unit B106 acquires the third rearrangement result π B Among (S), the eighth management information managed by the second device B<π B (S)> B (S203).

[0079] FIG. 11 is a diagram showing an example of the details of the processes of S200 to S203. The processes of S200 to S203 are executed using a secret sharing scheme. In this embodiment, a case where the secret sharing scheme described in "Wenjing Fang, Derun Zhao, Jin Tan, Chaochao Chen, Chaofan Yu, Li Wang, Lei Wang, Jun Zhou, Benyu Zhang, "Large-Scale Secure XGB for Vertical Federated Learning" is used is taken as an example. This method is also applied to the processes of S300 to S303 and S304 to S307 described later. Note that the processes of S200 to S203, S300 to S303, and S304 to S307 may be executed using other secret sharing schemes. Examples of other secret sharing schemes are as described above.

[0080] For example, the first key acquisition unit A111 acquires a first public key PK1 and a first secret key SK1 (S2000). In this embodiment, a case is exemplified in which the first key acquisition unit A111 acquires the first public key PK1 and the first secret key SK1 based on a public key cryptosystem (including a homomorphic public key cryptosystem). For example, various methods such as RSA, Paillier, DSA, or ECDSA may be used as the public key cryptosystem. Also, an encryption method other than the public key cryptosystem may be used. For example, the first key acquisition unit A111 may generate a key based on a common key cryptosystem, a secret key cryptosystem, or a hybrid method. When cryptographic communication is performed between the first device A and the second device B, any key in various known encryption methods may be used.

[0081] For example, the first public key transmission unit A112 transmits the first public key PK1 to the second device B (S2001). The first public key reception unit B115 of the second device B receives the first public key PK1 from the first device A (S2002). The transfer of the first public key PK1 itself may utilize a known public key cryptography method.

[0082] For example, the fifth management information acquisition unit A105 acquires the fifth management information based on the second random number. <s> A In S2003, the fifth management information acquisition unit A105 generates second random numbers the same number as the number of elements i, and acquires the fifth management information <s> A The fifth management information acquisition unit A105 acquires the fifth management information without using the second random number. <s> A For example, the fifth management information acquisition unit A105 may acquire the fifth management information including a predetermined value. <s> A The fifth management information may be acquired. <s> A may be determined based on the second random number.

[0083] For example, the seventh management information acquisition unit B105 acquires the seventh management information based on the third random number. <s> B In S2004, the seventh management information acquisition unit B105 generates third random numbers the same number as the number of elements i, and acquires the seventh management information <s> B The seventh management information acquisition unit B105 acquires the seventh management information without using the third random number. <s> B For example, the seventh management information acquisition unit B105 may acquire the seventh management information including a predetermined value. <s> B The seventh management information may be acquired. <s> B may be determined based on a third random number.

[0084] For example, the eighth management information acquisition unit B106 determines whether the eighth management information<π based on the fourth random number. B (S)> B In S2005, the eighth management information acquisition unit B106 generates fourth random numbers, the number of which is the same as the number of elements i, and obtains the eighth management information<π B (S)> B The eighth management information acquisition unit B106 acquires the eighth management information<π B (S)> B For example, the eighth management information acquisition unit B106 may acquire eighth management information<π B (S)> B Alternatively, the eighth management information <π B (S)> B Alternatively, only a part of the elements may be determined based on the fourth random number.

[0085] For example, the fifth management information transmission unit A110 transmits the fifth management information to the second device B. <s> A In S2006, the fifth management information transmission unit A110 transmits the fifth management information based on the first public key PK1 (S2006). <s> A and transmits the encrypted fifth management information [ <s> A In this embodiment, the encrypted information is enclosed in brackets [ ]. The fifth management information transmission unit A110 may perform encryption based on the first secret key SK. When an encryption method other than the public key encryption method is used, the fifth management information transmission unit A110 transmits to the second device B the fifth management information [ ] encrypted using the other encryption method. <s> A For example, when encryption communication is not used, the fifth management information transmission unit A110 may transmit the fifth management information in plain text to the second device B. <s> A may be transmitted.

[0086] For example, the fifth management information receiving unit B111 receives the fifth management information from the first device A. <s> A In S2007, the fifth management information receiving unit B111 of the second device B receives the encrypted fifth management information [ <s> A When an encryption method other than the public key encryption method is used, the fifth management information receiving unit B111 receives the fifth management information [ <s> A ] is also received in plain text. <s> A When the fifth management information is transmitted, the fifth management information receiving unit B111 receives the fifth management information in plain text. <s> A Receive.

[0087] For example, the third sorting result acquisition unit B112 acquires the fifth management information <s> A and the 7th Management Information <s> B Based on this, the third permutation result π B (S)= <s> A + <s> B (S2008). Note that the third rearrangement result acquisition unit B112 acquires the third rearrangement result π based on subtraction, multiplication, division, or other calculation formulas, rather than addition. B In S2008, the third sorting result acquisition unit B112 may acquire the seventh management information (S) based on the first public key PK1. <s> B The encrypted fifth management information [ <s> A ] and the encrypted seventh management information [ <s> B ] and the encrypted third permutation result [π B (S)]=[ <s> A ]+[ <s> B ] to obtain.

[0088] For example, the sixth management information acquisition unit B113 may obtain the third rearrangement result π B (S) and the 8th management information <π B (S)> B Based on the sixth management information <π B (S)> A = π B (S)-<π B (S)> B (S2009). Note that the sixth management information acquisition unit B113 acquires the sixth management information<π based on addition, multiplication, division, or other calculation formulas, not subtraction. B (S)> A In S2009, the sixth management information acquisition unit B113 may acquire the eighth management information<π based on the first public key PK1. B (S)> B Encrypt the third permutation result [π B (S)] and the encrypted eighth management information [<π B (S)> B ] and the encrypted sixth management information [<π B (S)> A ]=[π B (S)]-[<π B (S)> B The sixth management information acquisition unit B113 acquires the eighth management information<π B (S)> may be encrypted. In addition, when encryption communication is not used, the sixth management information acquisition unit B113 may encrypt the eighth management information <π B (S)> B In this case, encryption of the data may not be performed.

[0089] For example, the sixth management information transmission unit B114 transmits to the first device A the sixth management information<π B (S)> A In S2010, the sixth management information transmission unit B114 transmits the encrypted sixth management information [<π B (S)> A When encrypted communication is not used, the sixth management information transmission unit B114 transmits the sixth management information in plain text <π B (S)> encryption may not be performed.

[0090] For example, the first management information acquisition unit A103 acquires from the second device B the sixth management information<π B (S)> A and receiving the first management information <π B (π A (a))> A In S2011, the first management information acquisition unit A103 acquires the encrypted sixth management information [<π B (S)> A ] is received and decrypted with the first secret key SK1, and the first management information <π B (π A (a))> A The first management information acquisition unit A103 acquires the sixth management information [<π B (S)> A ] may be decrypted. Furthermore, if encrypted communication is not used, the first management information acquisition unit A103 does not need to execute decryption. The above-described processes of S2000 to S2011 are the details of the processes of S200 to S203.

[0091] Returning to FIG. 10, for example, the first public information acquisition unit B107 acquires the first rearrangement information π A A fourth permutation result π is a result of permutation of the first secret information a and is managed in a distributed manner by the first device A and the second device B. A In (a), the ninth management information managed by the second device B<π A (a)> B As a result, the first public information r to be made public to the first device A is acquired (S204). In S204, the first public information acquisition unit B107 of the second device B acquires the first public information r based on the first random number. The first public information acquisition unit B107 generates first random numbers the same number as the number of elements i based on any random number generation algorithm exemplified in this embodiment, and acquires them as the first public information r. The first public information acquisition unit B107 may acquire a predetermined fixed value as the first public information r instead of the first random number, or may acquire the current date and time as the first public information r.

[0092] For example, the first public information transmission unit B108 transmits the first public information r to the first device A (S205). In the present embodiment, an example is given in which the first public information transmission unit B108 encrypts the first public information r based on a predetermined encryption algorithm and then transmits it to the first device A, but the first public information transmission unit B108 may transmit the first public information r to the first device A in plain text without encrypting it.

[0093] For example, the first public information receiving unit A106 receives the first public information r from the second device B (S206). In this embodiment, since the first public information r is encrypted, the first public information receiving unit A106 decrypts the first public information r received from the second device B. When the first public information r is transmitted in plaintext, the first public information receiving unit A106 does not perform decryption.

[0094] For example, the tenth management information acquisition unit A107 acquires the fourth rearrangement result π A Among (a), the tenth management information managed by the first device A<π A (a)> A In S207, the tenth management information acquisition unit A107 acquires the fourth rearrangement result π A The difference between (a) and the first public information r is the tenth management information < π A (a)> A = π A (a)-r. Note that the tenth management information acquisition unit A107 acquires the tenth management information<π based on addition, multiplication, division, or other calculation formulas, not subtraction. A (a)> A may be obtained.

[0095] For example, the second public information acquisition unit A108 acquires the tenth management information<π A (a)> A and the fifth management information <s> A Based on the above, the second public information <π A (a)> A - <s> A (S208). Note that the second public information acquisition unit A108 may acquire the second public information based on addition, multiplication, division, or other calculation formulas instead of subtraction.

[0096] For example, the second public information transmission unit A109 transmits the second public information<π to the second device B. A (a)> A - <s> A In this embodiment, the second public information transmission unit A109 transmits the second public information<π based on a predetermined encryption algorithm. A (a)> A - <s> A The second public information transmission unit A109 transmits the second public information<π to the second device B as an example. A (a)> A - <s> A may be sent in the clear without encryption.

[0097] For example, the second public information receiving unit B109 receives the second public information <π A (a)> A - <s> A In this embodiment, the second public information<π is received (S210). A (a)> A - <s> A Since the second public information receiving unit B109 receives the second public information from the first device A, A (a)> A - <s> A Decrypt the second public information <π A (a)> A - <s> A If the second public information is transmitted in plain text, the second public information receiving unit B109 does not perform decryption.

[0098] For example, the first calculation information acquisition unit B110 acquires the eighth management information <s> B , 9th management information <π A (a)> B , and the second public information <π A (a)> A - <s> A Based on this, the first calculation information π A (a) -S=<π A (a)> A - <s> A +<π A (a)> B - <s> B (S211). The first calculation information acquisition unit B110 may execute the calculation in S211 based on a predetermined calculation formula, or may execute the calculation in S211 based on another calculation formula.

[0099] For example, the third management information acquisition unit B103 of the second device B acquires the second rearrangement information π B The first calculation information π A (a)-The fifth permutation result π after S is permuted B (π A (a)-S) and the eighth management information <π B (S)> B Based on the above, the third management information <π B (π A (a))> B = π B (π A (a)-S)+<π B (S)> B (S212). The third management information acquisition unit B103 may execute the calculation of S212 based on a predetermined formula, or may execute the calculation of S212 based on another formula. The above-described processes of S200 to S212 are the details of the processes of S20 and S21.

[0100] Returning to FIG. 9, for example, the second management information acquisition unit A104 acquires the first rearrangement information π A and second sorting information π B The second secret information b is rearranged as a second rearrangement result π B (π A (b)) the second management information managed by the first device A <d> A =<π B (π A (b))> A In this embodiment, the second rearrangement result π is obtained (S30). B (π A (b)) and the second management information <d> A Let us consider the case where is in vector form, and the second rearrangement result π B (π A (b)) and the second management information <d> A may be a number of numbers, a matrix, a vector, a data frame, or other information.

[0101] For example, the fourth management information acquisition unit B104 acquires the second rearrangement result π B (π A (b)) the fourth management information managed by the second device B <d> B =<π B (π A (b))> B In the processes of S30 and S31, the first device A and the second device B each obtain the first rearrangement information π A and second sorting information π B This is a process for sharing the second secret information b using a secret sharing scheme by using rearrangement of

[0102] 12 is a diagram showing an example of the details of the processes of S30 and S31. For example, the eleventh management information acquisition unit A113 acquires the eleventh management information managed by the first device A from the second distributed information R managed in a distributed manner by the first device A and the second device B. <r> A (S300). The second shared information R is information to be shared in the secret sharing scheme. In this embodiment, the second shared information R is a random number, but the second shared information R may be any information and is not limited to a random number. For example, the second shared information R may be a predetermined fixed value or the current time.

[0103] A detailed example of the process of S300 will be described later with reference to Fig. 13. Similarly, a detailed example of S301 to S303 will be described later with reference to Fig. 13. <r> A The process for acquiring the eleventh management information is not limited to the example shown in FIG. 13. The eleventh management information acquisition unit A113 may acquire the eleventh management information based on a secret sharing scheme other than that shown in FIG. <r> A The secret sharing scheme itself may be a known method, such as Shamir's Secret Sharing or Blakley's Secret Sharing. This also applies to the cases where any information is shared in the following description.

[0104] For example, the twelfth management information acquisition unit A114 acquires the second shared information R from the first rearrangement information π A A sixth sorted result π which is a result sorted by the first device A and the second device B and is managed in a distributed manner. A Among (R), the 12th management information managed by the first device A<π A (R)> A The first device A and the second device B actually obtain the sixth permutation result π A (R) is not obtained, but the 6th rearrangement result π A (R) shares only.

[0105] For example, the thirteenth management information acquisition unit B116 acquires the thirteenth management information managed by the second device B from the second distributed information R. <r> B (S302). The fourteenth management information acquisition unit B117 acquires the first rearrangement information π A The seventh permutation result π is a result of permutation of the second secret information b, and is managed in a distributed manner by the first device A and the second device B. A Among (b), the 14th management information managed by the second device B<π A (b)> B As a result, the sixth rearrangement result is π A Among (R), the 15th management information managed by the second device B<π A (R)> B (S303).

[0106] FIG. 13 is a diagram showing an example of the details of the processes of S300 to S303. The second key acquisition unit B122 acquires a second public key PK2 and a second secret key SK2 (S3000). In this embodiment, the second key acquisition unit B122 acquires the second public key PK2 and the second secret key SK2 based on the public key cryptosystem technique, but the key itself generated by the second key acquisition unit B122 may be any of various known keys. For example, the second key acquisition unit B122 may generate a key based on the common key cryptosystem technique. When cryptographic communication is performed between the first device A and the second device B, any key in any of various known cryptosystems may be used.

[0107] For example, the second public key transmission unit B123 transmits the second public key PK2 to the first device A (S3001). The second public key reception unit A124 receives the second public key PK2 from the second device B (S3002). The transfer of the second public key PK2 itself may utilize a known public key cryptography method.

[0108] For example, the eleventh management information acquisition unit A113 acquires the eleventh management information based on the fifth random number. <r> A In S3003, the eleventh management information acquisition unit A113 generates fifth random numbers, the number of which is the same as the number of elements i, and acquires the eleventh management information <r> A The eleventh management information acquisition unit A113 acquires the eleventh management information without using the fifth random number. <r> A For example, the eleventh management information acquisition unit A113 may acquire the eleventh management information including a predetermined value. <r> A The 11th management information may be acquired. <r> A Only a part of the elements may be determined based on the fifth random number.

[0109] For example, the 12th management information acquisition unit A114 determines whether the 12th management information<π based on the sixth random number. A (R)> A In S3004, the 12th management information acquisition unit A114 generates sixth random numbers, the number of which is the same as the number of elements i, and obtains the 12th management information<π A (R)> A The 12th management information acquisition unit A114 acquires the 12th management information<π A (R)> A For example, the 12th management information acquisition unit A114 may acquire 12th management information<π A (R)> A The seventh management information may be acquired. <s> B Only a part of the elements may be determined based on the sixth random number.

[0110] For example, the 13th management information acquisition unit B116 acquires the 13th management information based on the seventh random number. <r> B In S3005, the 13th management information acquisition unit B116 generates seventh random numbers, the number of which is the same as the number of elements i, and acquires the 13th management information <r> B The thirteenth management information acquisition unit B116 acquires the thirteenth management information without using the seventh random number. <r> B For example, the thirteenth management information acquisition unit B116 may acquire the thirteenth management information including a predetermined value. <r> B You may also obtain the 13th management information. <r> B Only a part of the elements may be determined based on the seventh random number.

[0111] For example, the 13th management information transmission unit transmits the 13th management information to the first device A. <r> B In S3006, the 13th management information transmission unit of the second device B transmits the 13th management information based on the second public key PK2 (S3006). <r> B and transmits the encrypted 13th management information [ <r> B The thirteenth management information transmission unit may perform encryption based on the first secret key SK. When an encryption method other than the public key encryption method is used, the thirteenth management information transmission unit transmits the thirteenth management information [ <r> B For example, when encrypted communication is not used, the thirteenth management information transmission unit may transmit the thirteenth management information in plain text to the first device A. <r> B may be transmitted.

[0112] For example, the thirteenth management information receiving unit A120 receives the thirteenth management information from the second device B. <r> B In S3007, the 13th management information receiving unit A120 receives the encrypted 13th management information [ <r> B In S3007, the thirteenth management information receiving unit A120 receives the encrypted thirteenth management information [ <r> B If an encryption method other than the public key encryption method is used, the 13th management information receiving unit A120 receives the 13th management information [ <r> B ] is also received in plain text. <r> B In the case where the thirteenth management information is transmitted, the thirteenth management information receiving unit A120 receives the thirteenth management information in the plain text. <r> B Receive.

[0113] For example, the sixth sorting result acquisition unit A121 acquires the eleventh management information <r> A and the 13th Management Information <r> B Based on the sixth permutation result π A (R) = <r> A + <r> B (S3008). Note that the sixth rearrangement result acquisition unit A121 acquires the sixth rearrangement result π based on subtraction, multiplication, division, or other calculation formulas, rather than addition. A In S3008, the sixth rearrangement result acquisition unit A121 may acquire the eleventh management information (R) based on the second public key PK2. <r> A The encrypted 13th management information [ <r> B ] and the encrypted 11th management information [ <r> A ] and the encrypted sixth permutation result [π A (R)]=[ <r> A ]+[ <r> B ] to obtain.

[0114] For example, the 15th management information acquisition unit A122 obtains the 6th sorting result π A (R) and 12th Management Information <π A (R)> A Based on the above, the 15th management information <π A (R)> B = π A (R)-<π A (R)> A (S3009). Note that the 15th management information acquisition unit A122 acquires the 15th management information<π based on addition, multiplication, division, or other calculation formulas, rather than subtraction. A (R)> B In S3009, the fifteenth management information acquisition unit A122 may acquire the twelfth management information<π based on the second public key PK2. A (R)> A Encrypt the sixth permutation result [π A (R)] and the encrypted 12th management information [<π A (R)> A ] and the encrypted 15th management information [<π A (R)> B ]=[π A (R)]+[<π A (R)> A The fifteenth management information acquisition unit A122 acquires the twelfth management information<π A (R)> A In addition, when encryption communication is not used, the fifteenth management information acquisition unit A122 may encrypt the twelfth management information <π A (R)> A In this case, encryption of the data may not be performed.

[0115] For example, the 15th management information transmission unit A123 transmits to the second device B the 15th management information<π A (R)> B In S3010, the 15th management information transmission unit A123 transmits the encrypted 15th management information [<π A (R)> B When encrypted communication is not used, the fifteenth management information transmission unit A123 transmits the fifteenth management information in plain text <π A (R)> B In this case, encryption of the data may not be performed.

[0116] For example, the 14th management information acquisition unit B117 of the second device B receives from the first device A the 15th management information<π A (R)> B Receive the 15th management information <π A (R)> B , 14th management information <π A (b)> B In S3011, the 14th management information acquisition unit B117 of the second device B acquires the encrypted 15th management information [<π A (R)> B ] is received and decrypted with the second secret key SK2, and the decrypted 15th management information [<π A (R)> B ] to the 14th management information <π A (b)> B The 14th management information acquisition unit B117 acquires the 15th management information [<π A (R)> B ] may be decrypted. Furthermore, if encrypted communication is not used, the fourteenth management information acquisition unit B117 does not need to execute decryption. The above-described processes of S3000 to S3011 are the details of the processes of S300 to S303.

[0117] 12, the processes of S304 to S307 are the same as those of S200 to S203. Since the processes of S200 to S203 have already been executed, the processes of S304 to S307 may be omitted. In addition, in S304, the second management information acquisition unit A104 acquires the first disperse information S from the second rearrangement information π B The third sorted result π is managed in a distributed manner by the first device A and the second device B. B Among (S), the sixth management information managed by the first device A<π B (S)> A , the second management information <π B (π A (b))> A Obtain as.

[0118] For example, the third public information acquisition unit B118 acquires the second secret information b and the thirteenth management information <r> B Third public information b- disclosed to the first device A based on <r> B (S308). Note that the third public information acquisition unit B118 acquires the third public information b- <r> B may be obtained.

[0119] For example, the third public information transmission unit B119 transmits the third public information b- <r> B In this embodiment, the third public information transmission unit B119 transmits the third public information b- <r> B The third public information transmission unit B119 transmits the third public information b- <r> B may be sent in the clear without encryption.

[0120] For example, the third public information receiving unit A115 receives the third public information b- <r> B In this embodiment, the third public information b- <r> B Since the third public information b- <r> B Decrypt the third public information b- <r> B If the third public information receiving unit A115 transmits the information in plain text, the third public information receiving unit A115 does not perform decryption.

[0121] For example, the second calculation information acquisition unit A116 acquires eleventh management information <r> A And the third public information b- <r> B Based on this, the second calculation information bR= <r> A +(b- <r> B ) (S311). Note that the second calculation information acquisition unit A116 may execute the process of S311 based on subtraction, multiplication, division, or another calculation formula instead of addition.

[0122] For example, the 16th management information acquisition unit A117 acquires the first rearrangement information π A The 8th rearrangement result π in which the second calculation information bR is rearranged B (π A (b)-S) and the 12th management information <π A (R)> A Based on this, the seventh rearrangement result π A Among (b), the 16th management information managed by the first device A<π A (b)> A = π A (bR)+<π A (R)> A (S312). The sixteenth management information acquisition unit A117 may execute the process of S312 based on subtraction, multiplication, division, or another calculation formula instead of addition.

[0123] For example, the fourth public information acquisition unit A118 acquires 16th management information<π A (b)> A and the fifth management information <s> A and the fourth public information <π A (b)> A - <s> A (S313). Note that the fourth public information acquisition unit A118 may acquire the fourth public information based on addition, multiplication, division, or other calculation formulas instead of subtraction.

[0124] For example, the fourth public information transmission unit A119 transmits the fourth public information<π to the second device B. A (b)> A - <s> A In this embodiment, the fourth public information transmission unit A119 transmits the fourth public information<π based on a predetermined encryption algorithm (S314). A (b)> A - <s> A The fourth public information transmission unit A119 transmits the fourth public information<π to the second device B as A (b)> A - <s> A may be sent in the clear without encryption.

[0125] For example, the fourth public information receiving unit B120 receives the fourth public information <π A (b)> A - <s> A In this embodiment, the fourth public information<π is received (S315). A (b)> A - <s> A Since the fourth public information receiving unit B120 receives the fourth public information from the first device A, A (b)> A - <s> A Decrypt the fourth public information <π A (b)> A - <s> A is transmitted in plain text, the fourth public information reception unit B120 does not perform decryption.

[0126] For example, the third calculation information acquisition unit B121 determines whether the 14th management information<π A (b)> B , 7th management information <s> B , and the fourth public information <π A (b)> A - <s> A Based on the third calculation information π A (b) -S=<π A (b)> B - <s> B +(<π A (b)> A - <s> A ) (S316). The third calculation information acquisition unit B121 may execute the calculation in S316 based on a predetermined calculation formula, or may execute the calculation in S316 based on another calculation formula.

[0127] For example, the fourth management information acquisition unit B104 of the second device B acquires the second rearrangement information π B The third calculation information π A (b)-The ninth rearrangement result π B (π A (b)-S) and the eighth management information <π B (S)> B Based on the above, the fourth management information <π B (π A (b))> B = π B (π A (b) -S) + < π B (S)> B (S317). The fourth management information acquisition unit B104 may execute the calculation of S317 based on a predetermined formula, or may execute the calculation of S317 based on another formula. The above-described processes of S300 to S317 are the details of the processes of S30 and S31.

[0128] Returning to FIG. 10, for example, the first management information acquisition unit A125 acquires the first management information <c> A and the second management information <d> A Based on the first permutation result (π B (π A (a)) and the second rearrangement result π B (π A (b)) and a distributed product e that is managed in a distributed manner by the first device A and the second device B. i =c i d i Among them, the first management area managed by the first device A <e i > A In S40, the first management area acquisition unit A125 acquires the first management area based on the BMT (Beaver Multiplication Triples) method. <e i > A In addition, the first management area acquisition unit A125 acquires the first management area based on a secure computation other than the BMT. <e i > A may be obtained.

[0129] For example, the second management information acquisition unit B124 acquires the third management information <c> B and the fourth management information <d> B Based on this, the variance product e i =c i d i Of these, the second management area managed by the second device B <e i > B In S41, the second management area acquisition unit B124 acquires the second management area based on the BMT technique. <e i > B In addition, the second management area acquisition unit B124 acquires the second management area based on a secure computation other than the BMT. <e i > B The secure computation is a technique for a device to perform computation without disclosing information. For example, the secure computation may utilize a secret sharing scheme, multi-party computation (MPC), homomorphic cryptography, reversible secure computation, or a trusted execution environment (TEE).

[0130] For example, the first total acquisition unit A126, the first management total <e i > A The first sum Σ i <e i > A The second total acquisition unit B125 acquires the second management total (S42). <e i > B The second sum Σ i <e i > B The second total transmission unit B126 transmits the second total Σ i <e i > B In this embodiment, the second sum transmission unit B126 transmits the second sum Σ i <e i > B The second total transmission unit B126 transmits the second total Σ i <e i > B may be sent in the clear without encryption.

[0131] For example, the second total receiving unit A127 receives the second total Σ i <e i > B In this embodiment, the second sum Σ i <e i > B Since the second total receiving unit A127 receives the second total Σ i <e i > B Decrypt the second sum Σ i <e i > B is transmitted in the clear, the second sum receiver A127 does not perform decryption.

[0132] The secret product calculation unit A128 calculates the first sum Σ i <e i > A And the second sum Σ i <e i > B Based on,and, the secret product Σ i a i b i =Σ i c i d i =Σ i e i =Σ i <e i > A +Σ i <e i > B (S46), and the process ends. i a i b i We will explain the case where the secret product Σ i a i b i As described above, in the calculation system 10, the first management information <c> A , the second management information <d> A , the third management information <c> B and fourth management information <d> B Based on the above, a secret product Σ i a i b i =Σ i c i d i =Σ i e i =Σ i <e i > A +Σ i <e i > B That is, in the calculation system 10, the first sum Σ i <e i > A And the second sum Σ i <e i > B Based on,and, the secret product Σ i a i b i is calculated.

[0133] For example, in the computing system 10, the secret product Σ i a i b i Based on this, learning of the learning model M in the federated learning is performed. For example, the learning model M may be a model in stacking learning that performs a predetermined estimation based on the estimation results of each of a plurality of other learning models. In this embodiment, the second device B performs learning of the secret product Σ i a i b i For example, the second device B performs learning of the second learning model BM based on the calculation formula shown in FIG.

[0134] [4. Summary of the embodiment] In this embodiment, the first device A processes the first sorting result (π B (π A Among (a), the first management information managed by the first device A <c> A =<π B (π A (a))> A The first device A obtains the second sorting result π B (π A (b)) the second management information managed by the first device A <d> A =<π B (π A (b))> A The second device B obtains the first sorting result π B (π A (a)) the third management information managed by the second device B <c> B =<π B (π A (a))> B The second device B obtains the second sorting result π B (π A (b)) the fourth management information managed by the second device B <d> B =<π B (π A (b))> B In the computing system 10, the first management information is obtained. <c> A , the second management information <d> A , the third management information <c> B and fourth management information <d> B Based on the above, a secret product Σ i a i b i is calculated. Since the first secret information a secretly managed by the first device A and the second secret information b secretly managed by the second device B are not directly transmitted, the secret product Σ i a i b i This improves security when calculating the first secret information a and the second secret information b. Even if a third party obtains information exchanged between the first device A and the second device B, the third party cannot identify the first secret information a and the second secret information b, so the risk of the first secret information a and the second secret information b being leaked to a third party can be reduced. In particular, compared to conventional secret sharing schemes, this scheme has a novel configuration, such as rearrangement of elements, which further improves security.

[0135] In addition, the first device A manages the fifth management information S managed by the first device A among the first distributed information S managed in a distributed manner by the first device A and the second device B. <s> A The first device A obtains the third sorted result π B Among (S), the sixth management information managed by the first device A<π B (S)> A , the first management information <π B (π A (a))> A The second device B acquires the seventh management information managed by the second device B from the first distributed information S as <s> B The second device B obtains the third permutation result π B Among (S), the eighth management information managed by the second device B<π B (S)> B The second device B obtains the fourth sorted result π A In (a), the ninth management information managed by the second device B<π A (a)> B The second device B transmits the first public information r to the first device A. The first device A receives the first public information r from the second device B. The first device A obtains the fourth rearrangement result π based on the first public information r. A Among (a), the tenth management information managed by the first device A<π A (a)> A The first device A obtains the tenth management information <π A (a)> A and the fifth management information <s> A Based on the above, the second public information <π A (a)> A - <s> A The first device A obtains the second public information <π A (a)> A - <s> A The second device B transmits the second public information <π A (a)> A - <s> A The second device B receives the eighth management information. <s> B , 9th management information <π A (a)> B , and the second public information <π A (a)> A - <s> A Based on this, the first calculation information π A (a) -S=<π A (a)> A - <s> A +<π A (a)> B - <s> B The second device B obtains the second rearrangement information π B The first calculation information π A (a)-The fifth permutation result π after S is permuted B (π A (a)-S) and the eighth management information <π B (S)> B Based on the above, the third management information <π B (π A (a))> B This obtains the first rearrangement information π A and second sorting information π B Since the first secret information a can be shared between the first device A and the second device B by using the rearrangement based on the above, security is improved compared to the conventional method in which rearrangement is not used.

[0136] Furthermore, the second device B acquires the first public information r based on the first random number. As a result, even if a third party acquires the first public information r transmitted to the network N, the first secret information a and the second secret information b cannot be identified by that alone, thereby reducing the risk that the first secret information a and the second secret information b will be leaked to a third party.

[0137] In addition, the first device A transmits fifth management information to the second device B. <s> A The second device B transmits the fifth management information from the first device A. <s> A The second device B receives the fifth management information. <s> A and the 7th Management Information <s> B Based on this, the third permutation result π B The second device B obtains the third sorting result π B (S) and the 8th management information <π B (S)> B Based on the sixth management information <π B (S)> A The second device B obtains the sixth management information<π B (S)> A The first device A transmits the sixth management information<π B (S)> A and receiving the first management information <π B (π A (a))> A As a result, if a third party were to acquire the fifth management information <s> A and 6th management information <π B (S)> A Even if the secret information a and the second secret information b are obtained, the first secret information a and the second secret information b cannot be identified by that alone, so the risk of the first secret information a and the second secret information b being leaked to a third party can be reduced.

[0138] The first device A also generates fifth management information based on the second random number. <s> A The second device B obtains the seventh management information based on the third random number. <s> B The second device B obtains the eighth management information<π based on the fourth random number. B (S)> B This allows a random number element to be included in the information available to a third party, thereby reducing the risk that the first secret information a and the second secret information b will be leaked to a third party.

[0139] The first device A also acquires a first public key PK1 and a first secret key SK1. The first device A transmits the first public key PK1 to the second device B. The second device B receives the first public key PK1 from the first device A. The first device A generates fifth management information based on the first public key PK1. <s> A and transmits the encrypted fifth management information [ <s> A The second device B transmits the encrypted fifth management information [ <s> A The second device B receives the seventh management information PK1 based on the first public key PK1. <s> B The encrypted fifth management information [ <s> A ] and the encrypted seventh management information [ <s> B ] and the encrypted third permutation result [π B (S)] based on the first public key PK1. B (S)> B Encrypt the third permutation result [π B (S)] and the encrypted eighth management information [<π B (S)> B ] and the encrypted sixth management information [<π B (S)> A The second device B transmits the encrypted sixth management information [<π B (S)> A The first device A transmits the encrypted sixth management information [<π B (S)> A ] is received and decrypted with the first secret key SK1, and the first management information <π B (π A (a))> A As a result, the exchange of information, which is already difficult for a third party to identify the first secret information a and the second secret information b even if they obtain it, is encrypted using the first public key PK1 and the first secret key SK1, thereby further enhancing security.

[0140] In addition, the first device A receives the eleventh management information R managed by the first device A from the second distributed information R managed in a distributed manner by the first device A and the second device B. <r> A The first device A obtains the sixth sorted result π A Among (R), the 12th management information managed by the first device A<π A (R)> A The first device A acquires the fifth management information S managed by the first device A from among the first distributed information S managed in a distributed manner by the first device A and the second device B. <s> A The first device A obtains the third sorted result π B Among (S), the sixth management information managed by the first device A<π B (S)> A , the second management information <π B (π A (b))> A The second device B acquires the thirteenth management information managed by the second device B from the second distributed information R as <r> B The second device B obtains the seventh sorted result π A Among (b), the 14th management information managed by the second device B<π A (b)> B As a result, the sixth rearrangement result is π A Among (R), the 15th management information managed by the second device B<π A (R)> B The second device B acquires the seventh management information managed by the second device B from the first distributed information S. <s> B and a seventh management information acquisition unit B105 that acquires the third rearrangement result π B Among (S), the eighth management information managed by the second device B<π B (S)> B The second device B obtains the second secret information b and the thirteenth management information <r> B Third public information b- disclosed to the first device A based on <r> B The second device B transmits the third public information b- <r> B The first device A transmits the third public information b- <r> B The first device A receives the eleventh management information. <r> A And the third public information b- <r> B and a second calculation information acquisition unit A116 that acquires second calculation information bR based on the first rearrangement information π A The 8th rearrangement result π in which the second calculation information bR is rearranged B (π A (b)-S) and the 12th management information <π A (R)> A Based on this, the seventh rearrangement result π A Among (b), the 16th management information managed by the first device A<π A (b)> A = π A (bR)+<π A (R)> A The first device A obtains the 16th management information <π A (b)> A and the fifth management information <s> A and the fourth public information <π A (b)> A - <s> A The first device A obtains the fourth public information <π A (b)> A - <s> A The second device B transmits the fourth public information <π A (b)> A - <s> A The second device B receives the 14th management information <π A (b)> B , 7th management information <s> B , and the fourth public information <π A (b)> A - <s> A Based on the third calculation information π A (b) -S=<π A (b)> B - <s> B +(<π A (b)> A - <s> A The second device B obtains the second rearrangement information π B The third calculation information π A (b)-The ninth rearrangement result π B (π A (b)-S) and the eighth management information <π B (S)> B Based on the above, the fourth management information <π B (π A (b))> B This obtains the first rearrangement information π A and second sorting information π B Since the second secret information b can be shared between the first device A and the second device B by using the rearrangement based on the above, security is improved compared to the conventional method in which rearrangement is not used.

[0141] The second device B also transmits thirteenth management information to the first device A. <r> B The first device A transmits the thirteenth management information from the second device B. <r> B The first device A receives the eleventh management information. <r> A and the 13th Management Information <r> B Based on the sixth permutation result π A (R). The first device A obtains the sixth permutation result π A (R) and 12th Management Information <π A (R)> A Based on the above, the 15th management information <π A (R)> B The first device A obtains the 15th management information<π A (R)> B The second device B transmits the 15th management information<π A (R)> B Receive the 15th management information <π A (R)> B , 14th management information <π A (b)> B As a result, if a third party were to obtain the 13th management information <r> B and 15 Management Information <π A (R)> B Even if the secret information a and the second secret information b are obtained, the first secret information a and the second secret information b cannot be identified by that alone, so the risk of the first secret information a and the second secret information b being leaked to a third party can be reduced.

[0142] In addition, the first device A generates eleventh management information based on the fifth random number. <r> A The first device A obtains the 12th management information<π based on the 6th random number. A (R)> A The second device B obtains the thirteenth management information based on the seventh random number. <r> B This allows a random number element to be included in the information available to a third party, thereby reducing the risk that the first secret information a and the second secret information b will be leaked to a third party.

[0143] Further, the second device B acquires a second public key PK2 and a second secret key SK2. The second device B transmits the second public key PK2 to the first device A. The first device A receives the second public key PK2 from the second device B. The second device B derives thirteenth management information based on the second public key PK2. <r> B and transmits the encrypted 13th management information [ <r> B The first device A transmits the encrypted 13th management information [ <r> B The first device A receives the eleventh management information PK1 based on the second public key PK2. <r> A The encrypted 13th management information [ <r> B ] and the encrypted 11th management information [ <r> A ] and the encrypted sixth permutation result [π A (R)]. The first device A obtains the 12th management information <π A (R)> A Encrypt the sixth permutation result [π A (R)] and the encrypted 12th management information [<π A (R)> A ] and the encrypted 15th management information [<π A (R)> B The first device A transmits the encrypted 15th management information [<π A (R)> B The 14th management information acquisition unit B117 of the second device B transmits the encrypted 15th management information [<π A (R)> B ] is received and decrypted with the second secret key SK2, and the decrypted 15th management information [<π A (R)> B ] to the 14th management information <π A (b)> B As a result, the exchange of information, which is already difficult for a third party to identify the first secret information a and the second secret information b even if they obtain it, is encrypted using the second public key PK2 and the second secret key SK2, thereby further enhancing security.

[0144] The first device A also stores first management information <c> A and the second management information <d> A Based on the first permutation result (π B (π A (a)) and the second rearrangement result π B (π A (b)) and a distributed product e that is managed in a distributed manner by the first device A and the second device B. i Among them, the first management area managed by the first device A <e i > A The first device A acquires the first management area. <e i > A The first sum Σ i <e i > A The second device B acquires the third management information. <c> B and the fourth management information <d> B Based on this, the variance product e i Of these, the second management area managed by the second device B <e i > B The second device B acquires the second management area. <e i > B The second sum Σ i <e i > B Obtain the first sum Σ i <e i > A And the second sum Σ i <e i > B Based on,and, the secret product Σ i a i b i is calculated. Even if a third party obtains information exchanged between the first device A and the second device B, the third party cannot identify the first secret information a and the second secret information b, so that the risk of the first secret information a and the second secret information b being leaked to a third party can be reduced.

[0145] In addition, the first device A calculates the first management area based on the BMT (Beaver Multiplication Triples) method. <e i > A The second device B obtains the second management area based on the BMT method. <e i > B This will obtain the first control product. <e i > A and the second management <e i > B This also uses secret computation techniques, further enhancing security.

[0146] In addition, the second device B is connected to the first device A via the second total Σ i <e i > B The first device A transmits the second sum Σ i <e i > B The first device A receives the first sum Σ i <e i > A And the second sum Σ i <e i > B Based on,and, the secret product Σ i a i b i This calculates the secret product Σ i a i b i Even if a third party obtains the information necessary for the calculation of (a), the first secret information a and the second secret information b cannot be identified by that alone, thereby reducing the risk that the first secret information a and the second secret information b will be leaked to a third party.

[0147] In addition, the first secret information a, the second secret information b, and the first sorting result (π B (π A (a)), 1st management information <c> A , the second rearrangement result π B (π A (b)), the second management information <d> A , the third management information <c> B , and the fourth management information <d> B Each of the first permutation information π A and second sorting information π B Each of is in array form. i a i b i This allows the computing system 10 to calculate the dot product of vectors without transmitting the vector, which should be kept secret, between the first device A and the second device B.

[0148] In addition, in the computation system 10, the secret product Σ i a i b i Based on this, the learning of the learning model in the federated learning is performed. This increases security during learning in the federated learning.

[0149] The first secret information is information about each of a plurality of users in the first service. The second secret information is information about each of a plurality of users in the second service. The calculation system 10 can calculate the product while keeping the information about the users secret, thereby reducing the risk of the information about the users being leaked.

[0150] [5. Modifications] The present disclosure is not limited to the above-described embodiment. It can be modified as appropriate without departing from the spirit of the present disclosure. For example, the calculation system 1 can be applied to situations other than federated learning. The calculation system 1 can be applied to situations in which a product of multiple pieces of secret information is calculated for some purpose.

[0151] For example, the functions described as being realized by the first device A or the second device B may be realized by at least one computer in the computing system 10, or the functions may be shared among multiple computers. In this case, each of the multiple computers may transmit its own processing results to the other computers, thereby sharing the functions.

[0152] [6. Notes] For example, the computing system may be configured as follows. (1) a first device that secretly manages first secret information having an element count of 2 or more and first rearrangement information for rearranging the elements of the element count; a second device that secretly manages second secret information of the number of elements and second rearrangement information for rearranging the elements of the number of elements; Including, The first device is a first management information acquisition unit that acquires first management information managed by the first device from a first permutation result that is a result of permutation of the first secret information by the first permutation information and the second permutation information and is managed in a distributed manner by the first device and the second device; a second management information acquisition unit that acquires second management information managed by the first device from a second sorting result that is a result of sorting the second secret information by the first sorting information and the second sorting information and is managed in a distributed manner by the first device and the second device; Including, The second device is a third management information acquisition unit that acquires third management information managed by the second device from the first sorting result; a fourth management information acquisition unit that acquires fourth management information managed by the second device from the second sorting result; Including, a secret product which is a product of the first secret information and the second secret information is calculated based on the first management information, the second management information, the third management information, and the fourth management information; Calculation system. (2) the first device further includes a fifth management information acquisition unit configured to acquire fifth management information managed by the first device from first distributed information managed in a distributed manner by the first device and the second device; the first management information acquisition unit of the first device acquires, as the first management information, sixth management information managed by the first device from among a third sorting result that is a result of sorting the first shared information by the second sorting information and is managed in a distributed manner by the first device and the second device; The second device is a seventh management information acquisition unit that acquires seventh management information managed by the second device from the first shared information; an eighth management information acquisition unit that acquires eighth management information managed by the second device from the third sorting result; a first public information acquisition unit that acquires first public information to be made public to the first device as ninth management information managed by the second device from among a fourth permutation result that is a result of permuting the first secret information with the first permutation information and is distributed and managed by the first device and the second device; a first public information transmission unit that transmits the first public information to the first device; Further comprising: The first device is a first public information receiving unit that receives the first public information from the second device; a tenth management information acquisition unit that acquires tenth management information managed by the first device from the fourth sorting result based on the first public information; a second public information acquisition unit that acquires second public information to be made public to the second device based on the tenth management information and the fifth management information; a second public information transmission unit that transmits the second public information to the second device; Further comprising: The second device is a second public information receiving unit that receives the second public information from the first device; a first calculation information acquisition unit that acquires first calculation information based on the eighth management information, the ninth management information, and the second public information; Further comprising: the third management information acquisition unit of the second device acquires the third management information based on a fifth sorting result in which the first calculation information is sorted by the second sorting information and the eighth management information; 10. A computing system as described in (1). (3) the first public information acquisition unit of the second device acquires the first public information based on a first random number; (2) A computing system according to claim 1. (4) the first device includes a fifth management information transmission unit configured to transmit the fifth management information to the second device; The second device is a fifth management information receiving unit that receives the fifth management information from the first device; the third sorting result acquisition unit acquiring the third sorting result based on the fifth management information and the seventh management information; a sixth management information acquisition unit that acquires the sixth management information based on the third sorting result and the eighth management information; a sixth management information transmitting unit configured to transmit the sixth management information to the first device; Further comprising: the first management information acquisition unit of the first device receives the sixth management information from the second device and acquires it as the first management information; A computing system according to (2) or (3). (5) the fifth management information acquisition unit of the first device acquires the fifth management information based on a second random number; the seventh management information acquisition unit of the second device acquires the seventh management information based on a third random number; the eighth management information acquisition unit of the second device acquires the eighth management information based on a fourth random number. (4) A computing system according to claim 1. (6) The first device is a first key acquisition unit that acquires a first public key and a first private key; a first public key transmission unit that transmits the first public key to the second device; Further comprising: the second device further includes a first public key receiving unit that receives the first public key from the first device; a fifth management information transmission unit of the first device encrypts the fifth management information based on the first public key, and transmits the encrypted fifth management information to the second device; the fifth management information receiving unit of the second device receives the encrypted fifth management information from the first device; the third permutation result acquisition unit of the second device encrypts the seventh management information based on the first public key, and acquires the encrypted third permutation result based on the encrypted fifth management information and the encrypted seventh management information; the sixth management information acquisition unit of the second device encrypts the eighth management information based on the first public key, and acquires the encrypted sixth management information based on the encrypted third permutation result and the encrypted eighth management information; the sixth management information transmission unit of the second device transmits the encrypted sixth management information to the first device; the first management information acquisition unit of the first device receives the encrypted sixth management information from the second device, decrypts the sixth management information with the first private key, and acquires the sixth management information as the first management information; A computing system according to (4) or (5). (7) The first device is an eleventh management information acquisition unit that acquires eleventh management information managed by the first device from second distributed information managed in a distributed manner by the first device and the second device; a 12th management information acquisition unit that acquires 12th management information managed by the first device from a sixth sorting result that is a result of sorting the second shared information by the first sorting information and that is managed in a distributed manner by the first device and the second device; a fifth management information acquisition unit that acquires fifth management information managed by the first device from first distributed information managed in a distributed manner by the first device and the second device; Further comprising: the second management information acquisition unit of the first device acquires, as the second management information, sixth management information managed by the first device from among a third sorting result that is a result of sorting the first shared information by the second sorting information and is managed in a distributed manner by the first device and the second device; The second device is a thirteenth management information acquisition unit that acquires thirteenth management information managed by the second device from the second shared information; a 14th management information acquisition unit that acquires 15th management information managed by the second device from the 6th permutation result as 14th management information managed by the second device from a 7th permutation result that is a result of permuting the second secret information with the first permutation information and is distributed and managed by the first device and the second device; a seventh management information acquisition unit that acquires seventh management information managed by the second device from the first distributed information S; an eighth management information acquisition unit that acquires eighth management information managed by the second device from the third sorting result; a third public information acquisition unit that acquires third public information to be made public to the first device based on the second secret information and the thirteenth management information; a third public information transmitting unit that transmits the third public information to the first device; Further comprising: The first device is a third public information receiving unit that receives the third public information from the second device; a second calculation information acquisition unit that acquires second calculation information based on the eleventh management information and the third public information; a 16th management information acquisition unit that acquires 16th management information managed by the first device from the 7th sorting result based on an 8th sorting result in which the 2nd calculation information is sorted by the 1st sorting information and the 12th management information; a fourth public information acquisition unit that acquires fourth public information to be made public to the second device based on the sixteenth management information and the fifth management information; a fourth public information transmitting unit that transmits the fourth public information to the second device; Further comprising: The second device is a fourth public information receiving unit that receives the fourth public information from the first device; a third calculation information acquisition unit that acquires third calculation information based on the fourteenth management information, the seventh management information, and the fourth public information; Further comprising: the fourth management information acquisition unit of the second device acquires the fourth management information based on a ninth sorting result obtained by sorting the third calculation information by the second sorting information and the eighth management information; A computation system according to any one of (1) to (6). (8) the second device includes a thirteenth management information transmission unit that transmits the thirteenth management information to the first device; The first device is a thirteenth management information receiving unit that receives the thirteenth management information from the second device; a sixth sorting result acquisition unit that acquires the sixth sorting result based on the eleventh management information and the thirteenth management information; a fifteenth management information acquisition unit that acquires the fifteenth management information based on the sixth sorting result and the twelfth management information; a fifteenth management information transmission unit that transmits the fifteenth management information to the second device; Further comprising: the 14th management information acquisition unit of the second device receives the 15th management information from the first device and acquires the 15th management information as the 14th management information. (7) A computing system according to (7). (9) the eleventh management information acquisition unit of the first device acquires the eleventh management information based on a fifth random number; the 12th management information acquisition unit of the first device acquires the 12th management information based on a sixth random number; the thirteenth management information acquisition unit of the second device acquires the thirteenth management information based on a seventh random number; (8) A computing system according to (8). (10) The second device is a second key acquisition unit that acquires a second public key and a second private key; a second public key transmission unit that transmits the second public key to the first device; Further comprising: the first device further includes a second public key receiving unit configured to receive the second public key from the second device; the thirteenth management information transmission unit of the second device encrypts the thirteenth management information based on the second public key, and transmits the encrypted thirteenth management information to the first device; the thirteenth management information receiving unit of the first device receives the encrypted thirteenth management information from the second device; the sixth permutation result acquisition unit of the first device encrypts the eleventh management information based on the second public key, and acquires the encrypted sixth permutation result based on the encrypted thirteenth management information and the encrypted eleventh management information; the 15th management information acquisition unit of the first device encrypts the 12th management information based on the second public key, and acquires the encrypted 15th management information based on the encrypted sixth permutation result and the encrypted 12th management information; the fifteenth management information transmission unit of the first device transmits the encrypted fifteenth management information to the second device; the 15th management information acquisition unit of the second device receives the encrypted 15th management information from the first device, decrypts the 15th management information with the second private key, and acquires the decrypted 15th management information as the 14th management information. A computing system according to (8) or (9). (11) the fifth management information acquisition unit of the first device acquires the fifth management information based on a second random number; the seventh management information acquisition unit of the second device acquires the seventh management information based on a third random number; the eighth management information acquisition unit of the second device acquires the eighth management information based on a fourth random number; A computation system according to any one of (7) to (10). (12) The first device is a first key acquisition unit that generates a first public key and a first private key; a first public key transmission unit that transmits the first public key to the second device; Further comprising: the second device further includes a first public key receiving unit that receives the first public key from the first device; a fifth management information transmission unit of the first device encrypts the fifth management information based on the first public key, and transmits the encrypted fifth management information to the second device; the fifth management information receiving unit of the second device receives the encrypted fifth management information from the first device; the third permutation result acquisition unit of the second device encrypts the seventh management information based on the first public key, and acquires the encrypted third permutation result based on the encrypted fifth management information and the encrypted seventh management information; the sixth management information acquisition unit of the second device encrypts the eighth management information based on the first public key, and acquires the encrypted sixth management information based on the encrypted third permutation result and the encrypted eighth management information; the sixth management information transmission unit of the second device transmits the encrypted sixth management information to the first device; the first management information acquisition unit of the first device receives the encrypted sixth management information from the second device and decrypts it with the first private key; A computation system according to any one of (7) to (11). (13) The first device is a first management product acquisition unit that acquires a first management product managed by the first device from a distributed product that is a product of the first sorting result and the second sorting result based on the first management information and the second management information and is distributed and managed by the first device and the second device; a first total acquisition unit that acquires a first total that is a total of the first management products; Further comprising: The second device is a second management area acquisition unit that acquires a second management area managed by the second device from among the distribution areas based on the third management information and the fourth management information; a second total acquisition unit that acquires a second total that is a total of each element of the second management product; Further comprising: the secret product is calculated based on the first sum and the second sum; A computation system according to any one of (1) to (12). (14) the first management product acquisition unit of the first device acquires a first management product based on a Beaver Multiplication Triples (BMT) technique; the second management area acquisition unit of the second device acquires the second management area based on the BMT technique. (13) A computing system according to (13). (15) the second device further includes a second total transmitter configured to transmit the second total to the first device; The first device is a second total receiving unit that receives the second total from the second device; a secret product calculation unit that calculates the secret product based on the first sum and the second sum; The computing system according to (13) or (14), further comprising: (16) each of the first secret information, the second secret information, the first permutation result, the first management information, the second permutation result, the second management information, the third management information, and the fourth management information is in a vector format; Each of the first rearrangement information and the second rearrangement information is in an array format; The secret product is an inner product. A computation system according to any one of (1) to (15). (17) In the computing system, learning of a learning model in federated learning is performed based on the secret product. A computation system according to any one of (1) to (16). (18) the first secret information is information regarding each of a plurality of users in a first service; The second secret information is information regarding each of the plurality of users in the second service. A computation system according to any one of (1) to (17). [Explanation of symbols]

[0153] 10 Computing system, A first device, B second device, A11 control unit, A12 memory unit, A13 communication unit, B11 control unit, B12 memory unit, B13 communication unit, A100 first data memory unit, A101 first secret information acquisition unit, A102 first rearrangement information acquisition unit, A103 first management information acquisition unit, A104 second management information acquisition unit, A105 fifth management information acquisition unit, A106 first public information reception unit, A107 tenth management information acquisition unit, A108 second public information acquisition unit, A109 second public information transmission unit, A110 fifth management information transmission unit, A111 first key acquisition unit, A112 first public key transmission unit, A113 eleventh management information acquisition unit, A114 twelfth management information acquisition unit, A115 third public information reception unit, A116 Second calculation information acquisition unit, A117 Sixteenth management information acquisition unit, A118 Fourth public information acquisition unit, A119 Fourth public information transmission unit, A120 Thirteenth management information reception unit, A121 Sixth rearrangement result acquisition unit, A122 Fifteenth management information acquisition unit, A123 Fifteenth management information transmission unit, A124 Second public key reception unit, A125 First management product acquisition unit, A126 First total acquisition unit, A127 Second total reception unit, A128 Secret product calculation unit, B100 Second data storage unit, B101 Second secret information acquisition unit, B102 Second rearrangement information acquisition unit, B103 Third management information acquisition unit, B104 Fourth management information acquisition unit, B105 Seventh management information acquisition unit, B106 Eighth management information acquisition unit, B107 First public information acquisition unit, B108 First public information transmission unit, B109 second public information receiving unit, B110 first calculation information acquiring unit, B111 fifth management information receiving unit, B112 third sorting result acquiring unit, B113 sixth management information acquiring unit, B114 sixth management information transmitting unit, B115 first public key receiving unit, B116 thirteenth management information acquiring unit, B117 fourteenth management information acquiring unit, B118 third public information acquiring unit, B119 third public information transmitting unit, B120 fourth public information receiving unit, B121 third calculation information acquiring unit, B122 second key acquiring unit, B123 second public key transmitting unit, B124 second management product acquiring unit, B125 second total acquiring unit, B126 second total transmitting unit.< / d> < / c> < / d> < / c> < / d> < / c> < / d> < / c> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / r> < / r> < / r> < / r> < / r> < / r> < / s> < / r> < / s> < / r> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / d> < / c> < / d> < / c> < / d> < / c> < / d> < / c> < / d> < / c> < / d> < / c> < / d> < / c> < / d> < / c> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / s> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / r> < / d> < / d> < / d> < / d> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / s> < / c> < / c> < / c> < / c>

Claims

1. a first device that secretly manages first secret information having a number of elements equal to or greater than 2 and first rearrangement information for rearranging the elements of the number of elements; a second device that secretly manages second secret information of the number of elements and second rearrangement information for rearranging the elements of the number of elements; Including, The first device is a first management information acquisition unit configured to acquire first management information managed by the first device from a first permutation result that is a result of permutation of the first secret information by the first permutation information and the second permutation information and is managed in a distributed manner by the first device and the second device; a second management information acquisition unit configured to acquire second management information managed by the first device from a second sorting result that is a result of sorting the second secret information by the first sorting information and the second sorting information and is managed in a distributed manner by the first device and the second device; Including, The second device is a third management information acquisition unit that acquires third management information managed by the second device from the first sorting result; a fourth management information acquisition unit that acquires fourth management information managed by the second device from the second sorting result; Including, a secret product which is a product of the first secret information and the second secret information is calculated based on the first management information, the second management information, the third management information, and the fourth management information; Calculation system.

2. the first device further includes a fifth management information acquisition unit configured to acquire fifth management information managed by the first device from first distributed information managed in a distributed manner by the first device and the second device; the first management information acquisition unit of the first device acquires, as the first management information, sixth management information managed by the first device from a third sorting result that is a result of sorting the first shared information by the second sorting information and is managed in a distributed manner by the first device and the second device; The second device is a seventh management information acquisition unit that acquires seventh management information managed by the second device from the first shared information; an eighth management information acquisition unit that acquires eighth management information managed by the second device from the third sorting result; a first public information acquisition unit that acquires first public information to be made public to the first device as ninth management information managed by the second device from among a fourth sorting result that is a result of sorting the first secret information by the first sorting information and is managed in a distributed manner by the first device and the second device; a first public information transmission unit that transmits the first public information to the first device; Further comprising: The first device is a first public information receiving unit that receives the first public information from the second device; a tenth management information acquisition unit that acquires tenth management information managed by the first device from the fourth sorting result based on the first public information; a second public information acquisition unit that acquires second public information to be made public to the second device based on the tenth management information and the fifth management information; a second public information transmission unit that transmits the second public information to the second device; Further comprising: The second device is a second public information receiving unit that receives the second public information from the first device; a first calculation information acquisition unit that acquires first calculation information based on the eighth management information, the ninth management information, and the second public information; Further comprising: the third management information acquisition unit of the second device acquires the third management information based on a fifth sorting result in which the first calculation information is sorted by the second sorting information and the eighth management information; The computing system of claim 1 .

3. The first public information acquisition unit of the second device acquires the first public information based on a first random number. The computing system of claim 2 .

4. the first device includes a fifth management information transmission unit configured to transmit the fifth management information to the second device; The second device is a fifth management information receiving unit that receives the fifth management information from the first device; the third sorting result acquisition unit acquiring the third sorting result based on the fifth management information and the seventh management information; a sixth management information acquisition unit that acquires the sixth management information based on the third sorting result and the eighth management information; a sixth management information transmitting unit configured to transmit the sixth management information to the first device; Further comprising: the first management information acquisition unit of the first device receives the sixth management information from the second device and acquires it as the first management information; A computing system according to claim 2 or 3.

5. the fifth management information acquisition unit of the first device acquires the fifth management information based on a second random number; the seventh management information acquisition unit of the second device acquires the seventh management information based on a third random number; the eighth management information acquisition unit of the second device acquires the eighth management information based on a fourth random number; The computing system of claim 4.

6. The first device is a first key acquisition unit that acquires a first public key and a first private key; a first public key transmission unit that transmits the first public key to the second device; Further comprising: the second device further includes a first public key receiving unit configured to receive the first public key from the first device; a fifth management information transmission unit of the first device encrypts the fifth management information based on the first public key, and transmits the encrypted fifth management information to the second device; the fifth management information receiving unit of the second device receives the encrypted fifth management information from the first device; the third permutation result acquisition unit of the second device encrypts the seventh management information based on the first public key, and acquires the encrypted third permutation result based on the encrypted fifth management information and the encrypted seventh management information; the sixth management information acquisition unit of the second device encrypts the eighth management information based on the first public key, and acquires the encrypted sixth management information based on the encrypted third permutation result and the encrypted eighth management information; the sixth management information transmission unit of the second device transmits the encrypted sixth management information to the first device; the first management information acquisition unit of the first device receives the encrypted sixth management information from the second device, decrypts the sixth management information with the first private key, and acquires the sixth management information as the first management information; The computing system of claim 4.

7. The first device is an eleventh management information acquisition unit that acquires eleventh management information managed by the first device from second distributed information managed in a distributed manner by the first device and the second device; a twelfth management information acquisition unit that acquires twelfth management information managed by the first device from a sixth rearrangement result that is a result of rearrangement of the second shared information by the first rearrangement information and that is managed in a distributed manner by the first device and the second device; a fifth management information acquisition unit that acquires fifth management information managed by the first device from first distributed information managed in a distributed manner by the first device and the second device; Further comprising: the second management information acquisition unit of the first device acquires, as the second management information, sixth management information managed by the first device from a third sorting result that is a result of sorting the first shared information by the second sorting information and is managed in a distributed manner by the first device and the second device; The second device is a thirteenth management information acquisition unit that acquires thirteenth management information managed by the second device from the second shared information; a 14th management information acquisition unit that acquires 15th management information managed by the second device from the 6th sorting result as 14th management information managed by the second device from a 7th sorting result that is a result of sorting the second secret information by the 1st sorting information and that is managed in a distributed manner by the first device and the second device; a seventh management information acquisition unit that acquires seventh management information managed by the second device from the first distributed information S; an eighth management information acquisition unit that acquires eighth management information managed by the second device from the third sorting result; a third public information acquisition unit that acquires third public information to be made public to the first device based on the second secret information and the thirteenth management information; a third public information transmission unit that transmits the third public information to the first device; Further comprising: The first device is a third public information receiving unit that receives the third public information from the second device; a second calculation information acquisition unit that acquires second calculation information based on the eleventh management information and the third public information; a 16th management information acquisition unit that acquires 16th management information managed by the first device from the 7th sorting result based on an 8th sorting result in which the 2nd calculation information is sorted by the 1st sorting information and the 12th management information; a fourth public information acquisition unit that acquires fourth public information to be made public to the second device based on the sixteenth management information and the fifth management information; a fourth public information transmission unit that transmits the fourth public information to the second device; Further comprising: The second device is a fourth public information receiving unit that receives the fourth public information from the first device; a third calculation information acquisition unit that acquires third calculation information based on the fourteenth management information, the seventh management information, and the fourth public information; Further comprising: the fourth management information acquisition unit of the second device acquires the fourth management information based on a ninth sorting result in which the third calculation information is sorted by the second sorting information and the eighth management information; A computing system according to any one of claims 1 to 3.

8. the second device includes a thirteenth management information transmission unit configured to transmit the thirteenth management information to the first device; The first device is a thirteenth management information receiving unit that receives the thirteenth management information from the second device; a sixth sorting result acquisition unit that acquires the sixth sorting result based on the eleventh management information and the thirteenth management information; a fifteenth management information acquisition unit that acquires the fifteenth management information based on the sixth rearrangement result and the twelfth management information; a fifteenth management information transmitting unit configured to transmit the fifteenth management information to the second device; Further comprising: the fourteenth management information acquisition unit of the second device receives the fifteenth management information from the first device and acquires the fifteenth management information as the fourteenth management information; The computing system of claim 7.

9. the eleventh management information acquisition unit of the first device acquires the eleventh management information based on a fifth random number; the twelfth management information acquisition unit of the first device acquires the twelfth management information based on a sixth random number; the thirteenth management information acquisition unit of the second device acquires the thirteenth management information based on a seventh random number; The computing system of claim 8.

10. The second device is a second key acquisition unit that acquires a second public key and a second private key; a second public key transmission unit that transmits the second public key to the first device; Further comprising: the first device further includes a second public key receiving unit configured to receive the second public key from the second device; the thirteenth management information transmission unit of the second device encrypts the thirteenth management information based on the second public key, and transmits the encrypted thirteenth management information to the first device; the thirteenth management information receiving unit of the first device receives the encrypted thirteenth management information from the second device; the sixth permutation result acquisition unit of the first device encrypts the eleventh management information based on the second public key, and acquires the encrypted sixth permutation result based on the encrypted thirteenth management information and the encrypted eleventh management information; the fifteenth management information acquisition unit of the first device encrypts the twelfth management information based on the second public key, and acquires the encrypted fifteenth management information based on the encrypted sixth permutation result and the encrypted twelfth management information; the fifteenth management information transmission unit of the first device transmits the encrypted fifteenth management information to the second device; the fifteenth management information acquisition unit of the second device receives the encrypted fifteenth management information from the first device, decrypts the fifteenth management information with the second private key, and acquires the decrypted fifteenth management information as the fourteenth management information; The computing system of claim 8.

11. the fifth management information acquisition unit of the first device acquires the fifth management information based on a second random number; the seventh management information acquisition unit of the second device acquires the seventh management information based on a third random number; the eighth management information acquisition unit of the second device acquires the eighth management information based on a fourth random number; The computing system of claim 7.

12. The first device is a first key acquisition unit that generates a first public key and a first secret key; a first public key transmission unit that transmits the first public key to the second device; Further comprising: the second device further includes a first public key receiving unit configured to receive the first public key from the first device; a fifth management information transmission unit of the first device encrypts the fifth management information based on the first public key, and transmits the encrypted fifth management information to the second device; the fifth management information receiving unit of the second device receives the encrypted fifth management information from the first device; the third permutation result acquisition unit of the second device encrypts the seventh management information based on the first public key, and acquires the encrypted third permutation result based on the encrypted fifth management information and the encrypted seventh management information; the sixth management information acquisition unit of the second device encrypts the eighth management information based on the first public key, and acquires the encrypted sixth management information based on the encrypted third permutation result and the encrypted eighth management information; the sixth management information transmission unit of the second device transmits the encrypted sixth management information to the first device; the first management information acquisition unit of the first device receives the encrypted sixth management information from the second device and decrypts it with the first private key; The computing system of claim 7.

13. The first device is a first management product acquisition unit that acquires a first management product managed by the first device from a distributed product of the first sorting result and the second sorting result, the distributed product being managed in a distributed manner by the first device and the second device, based on the first management information and the second management information; a first total acquisition unit that acquires a first total that is a total of the first management products; Further comprising: The second device is a second management product acquisition unit that acquires a second management product managed by the second device from among the distribution products based on the third management information and the fourth management information; a second total acquisition unit that acquires a second total that is a total of each element of the second management product; Further comprising: the secret product is calculated based on the first sum and the second sum; A computing system according to any one of claims 1 to 3.

14. the first management product acquisition unit of the first device acquires a first management product based on a Beaver Multiplication Triples (BMT) technique; the second management area acquisition unit of the second device acquires the second management area based on the BMT technique.

14. The computing system of claim 13.

15. the second device further includes a second total transmitter configured to transmit the second total to the first device; The first device is a second total receiving unit for receiving the second total from the second device; a secret product calculation unit that calculates the secret product based on the first sum and the second sum; The computing system of claim 13 further comprising:

16. each of the first secret information, the second secret information, the first permutation result, the first management information, the second permutation result, the second management information, the third management information, and the fourth management information is in a vector format; Each of the first rearrangement information and the second rearrangement information is in an array format; The secret product is an inner product. A computing system according to any one of claims 1 to 3.

17. In the computing system, learning of a learning model in federated learning is performed based on the secret product. A computing system according to any one of claims 1 to 3.

18. the first secret information is information regarding each of a plurality of users in a first service; The second secret information is information regarding each of the plurality of users in the second service. A computing system according to any one of claims 1 to 3.

19. a first device that secretly manages first secret information having a number of elements equal to or greater than 2 and first rearrangement information for rearranging the elements of the number of elements; a second device that secretly manages second secret information of the number of elements and second rearrangement information for rearranging the elements of the number of elements; It is executed by The first device is acquiring first management information managed by the first device from a first permutation result, the first management information being a result of permutation of the first secret information by the first permutation information and the second permutation information and being managed in a distributed manner by the first device and the second device; acquiring second management information managed by the first device from a second sorting result, the second sorting result being a result of sorting the second secret information by the first sorting information and the second sorting information and managed in a distributed manner by the first device and the second device; The second device is obtaining third management information managed by the second device from the first sorting result; obtaining fourth management information managed by the second device from the second sorting result; a secret product which is a product of the first secret information and the second secret information is calculated based on the first management information, the second management information, the third management information, and the fourth management information; Calculation method.

20. A program for causing a computer to operate as the first device or the second device according to claim 1.