Secure calculation client device and attribute selection method

Abstract

A secure calculation client device according to the present invention comprises a secure calculation client unit, an attribute selection unit, and an AutoML unit. The secure calculation client unit is connected to a secure calculation system formed from a plurality of secure calculation servers, transmits, to the secure calculation system, a share of indication information pertaining to machine learning by an AI model, and receives a share of the evaluation values of the machine learning by the AI model. The attribute selection unit selects, as a selected attribute, an attribute to be used for the machine learning by the AI model, said selection being performed on the basis of the correlation of the attribute with an objective variable, and designates the selected attribute for the secure calculation client unit. The AutoML unit evaluates the AI model and a hyperparameter in accordance with a plaintext performance index based on the evaluation values.

Description

Secure computation client device and attribute selection method

[0001] The present invention relates to a secure computation client device connected to a secure computation system consisting of a plurality of secure computation servers, and an attribute selection method using the secure computation client device.

[0002] While the momentum for data utilization using AI and other technologies is growing, demand for privacy protection is also increasing, and there is a demand for an environment where highly confidential data can be analyzed using proven AI algorithms in fields such as medical care. Secure computing AI is a technology that brings together data accumulated by multiple companies and safely utilizes it. It enables the creation of models and inference (data operation in which only the calculation results are visible to anyone) using extremely secure and diverse AI algorithms while keeping the data confidential (secret distribution) and never restoring it. Furthermore, AutoML (Automated Machine Learning), as shown in Non-Patent Documents 1 and 2, is known as a technology for automatically tuning AI models.

[0003] Yao, Quanming & Wang, Mengshuo & Escalante, Hugo Jair & Isabelle, Guyon & Hu, Yi-Qi & Yu-Feng, Li & Tu, Wei-Wei & Qiang, Yang & Yang, Yu. (2018), "Taking Human out of Learning Applications: A Survey on Automated Machine Learning", 31, October, 2018, [Retrieved June 11, 2020], Internet<https: / / ar5iv.labs.arxiv.org / html / 1810.13306> Xin He, Kaiyong Zhao, Xiaowen Chu, "AutoML: A survey of the state-of-the-art", Knowledge-Based Systems, Volume 212, 2021, 106622, ISSN 0950-7051, [Retrieved June 11, 2024], Internet<https: / / doi.org / 10.1016 / j.knosys.2020.106622> .

[0004] However, to perform secure computation on machine learning in actual analyses and services, it is necessary to appropriately determine hyperparameters, such as the type of learning and the learning method to be used. However, this is less efficient than plaintext, and manually performing this is not practical. AutoML (Non-Patent Documents 1 and 2), known as a technology for automatically tuning AI models, automatically tunes by repeating various machine learning procedures multiple times, which results in computational costs that are too high for direct implementation using secure computation. To make AutoML more efficient, it is sufficient to reduce the number of attributes to be learned, but sequential search, simulated annealing, genetic algorithms, and stepwise methods require many repetitions and are therefore not efficient. Therefore, an efficient attribute selection method is needed. The present invention aims to provide an efficient attribute selection method.

[0005] The secure computation client device of the present invention comprises a secure computation client unit, an attribute selection unit, and an AutoML unit. The secure computation client unit is connected to a secure computation system consisting of one or more secure computation servers, and transmits shares of instruction information related to machine learning of an AI model to the secure computation system, and receives shares of evaluation values ​​of the machine learning of the AI ​​model. The attribute selection unit selects attributes to be used in machine learning of the AI ​​model based on correlation with the objective variable, sets them as selected attributes, and instructs the secure computation client unit to select them. The AutoML unit evaluates the AI ​​model and hyperparameters according to plaintext performance indicators based on the evaluation values.

[0006] According to the secure computation client device of the present invention, learning of an AI model is performed using secure computation, and the AutoML unit selects attributes to be used for learning based on performance indicators, which are plaintext statistics, so that an optimal machine learning model can be automatically searched for while maintaining data confidentiality. In addition, an efficient method for selecting attributes can be provided.

[0007] A secure computation system including a plurality of secure computation servers connected via a network, and a secure computation client device. A secure computation system including a plurality of secure computation servers connected via a network, and a secure computation client device. A secure computation client device including a plurality of secure computation servers connected via a network ...

[0008] Prior to describing the embodiments, the basic technical concepts used in the present disclosure will be described.

[0009] [Secret Sharing Technology] Secret sharing is a technology that converts data into multiple shares, allowing the original data to be restored using a certain number of shares or more, but making it impossible to restore the original data using shares less than the certain number. (k,n)-secret sharing, a type of secret sharing, divides an input plaintext into n shares, distributes the shares to n computing entities, and allows the plaintext to be restored using any k shares, but no information about the plaintext can be obtained using shares less than k. Here, n and k are integers greater than or equal to 1, and n≧k. A typical example of (k,n)-secret sharing is Shamir secret sharing, described in "A. Shamir, "How to share a secret," Communications of the ACM, Volume 22, Issue 11, pp. 612-613, 1979 (Reference 1)." The secret sharing used in this disclosure may be any method that can utilize secure computation, as described below.

[0010] [Secure computation technology] Secure computation is a technology in which data to be computed is secretly shared and stored among multiple computing entities, and the computing entities cooperate with each other to compute the shared values ​​of a function of the original data without restoring the original data.Secure computation uses secret sharing as an elemental technology.

[0011] The secure computation used in the first embodiment of the present disclosure may be one that allows various computations required for desired data processing to be performed on shared values ​​obtained by a specific secret sharing method. A secure computation technique that performs basic computations such as addition and multiplication on secret shared values ​​is described, for example, in "Koji Senda, Hiroki Hamada, Dai Igarashi, and Katsumi Takahashi, 'Rethinking Lightly Verifiable Three-Party Secure Function Computation,' Computer Security Symposium 2010, 2010 (Reference 2)." A secret matching technique that searches for information from secret shared values ​​of a data string while keeping it secret is described, for example, in "Koji Senda, Masayuki Terada, Takayasu Yamaguchi, Dai Igarashi, Hiroki Hamada, and Katsumi Takahashi, 'Secure Matching Protocol Considering Statistical Disclosure Control,' Information Processing Society of Japan Research Report, 2011-CSEC-52(12), 2011 (Reference 3)." A secret sorting technique for sorting a data string while keeping the secret sharing values ​​of the data string secret is described, for example, in "Hiroki Hamada, Dai Igarashi, Koji Senda, Katsumi Takahashi, "Linear Time Sorting on Secure Function Computation," Computer Security Symposium 2011, 2011 (Reference 4)."

[0012] Hereinafter, embodiments of the present invention will be described in detail. Components having the same functions are given the same numbers, and duplicated explanations will be omitted. First embodiment

[0013] FIG. 1 shows an example of the configuration of a secure computation system and a secure computation client device, each consisting of multiple secure computation servers connected via a network. FIG. 2 shows the processing flow of an attribute selection method in the secure computation client device. The secure computation system 10 is composed of X secure computation servers 100-1, ..., X. X may be, for example, 3 or greater, but may be any number greater than or equal to 1 as long as the method allows for secure computation, as described below. The secure computation system 10 can use existing technology (such as the technology disclosed in WO2012 / 046692 (family U.S. Patent US8,989,391B2) (Reference 5)). For example, when X = 3, one piece of data is converted into three shares, and three secure computation servers 100-1, 100-2, and 100-3 each record one share. Two shares are required to restore the data. In the secure computation system 10, calculations can be performed while maintaining the data in a secretly shared state (anonymized state), and the results are also recorded as anonymous shares on each secure computation server 100-x. If the result is a statistic, the statistic can be known by obtaining and restoring two or more shares, while maintaining the confidentiality of the original data.

[0014] The secure computation client device 200 includes a secure computation client unit 210, an attribute selection unit 230, an AutoML unit 220, a communication unit 280, and a recording unit 290. The communication unit 280 may be connected to a user device 300. In this case, encryption technology may be used for communication between the communication unit 280 and the user device 300. The recording unit 290 records records used for training the AI ​​model. Each record is composed of data of multiple attributes. The attributes include, for example, age, gender, basic medical information (such as the values ​​of blood components), and disease name.

[0015] The secure computation client unit 210 is connected to a secure computation system consisting of multiple secure computation servers 100-1, ..., X, and includes a secret sharing unit 211 and a restoration unit 212. The secure computation client unit 210 transmits shares of instruction information related to machine learning of the AI ​​model to the secure computation system 10, and receives shares of evaluation values ​​of the machine learning of the AI ​​model from the secure computation system 10. Here, "transmitting shares to the secure computation system 10" means transmitting shares corresponding to each secure computation server 100-x, and "receiving shares from the secure computation system 10" means receiving shares recorded by each secure computation server 100-x. In other words, the secret sharing unit 211 converts records, instruction information, etc. into multiple shares and transmits shares corresponding to each secure computation server 100-x. The instruction information is information such as which attribute is used to perform machine learning of the AI ​​model. The restoration unit 212 receives shares of calculation results and shares of evaluation values ​​in the secure computation system 10, and acquires plaintext calculation results and evaluation values. The evaluation value of the plaintext itself may be used as the performance index, or the evaluation value may be processed in some way and the result may be used as the performance index.

[0016] The secret sharing unit 211 converts a record to be used for machine learning of the AI ​​model into a plurality of shares in advance and transmits the corresponding shares to each secure computation server 100-x. Each secure computation server 100-x records the shares of data of attributes that make up the record.

[0017] The attribute selection unit 230 selects attributes to be used in machine learning of the AI ​​model based on their correlation with the objective variable, sets them as selected attributes, and instructs the secure computation client unit 210 (S231). When selecting attributes, if the attribute selection unit 230 uses a method that multiplies the L1 norm of the regression coefficient of the attribute as a penalty (e.g., LASSO regression or Ridge regression), the coefficient of an attribute that has a low correlation with the objective variable will ultimately become 0, making it possible to eliminate unrelated attributes. Alternatively, a correlation coefficient may be used to select attributes that have a high correlation with the objective variable. When using a correlation coefficient, since Pearson's correlation coefficient only corresponds to linearity, Spearman's or Kendall's correlation coefficient may be used.

[0018] The secret sharing unit 211 converts instruction information related to machine learning of the AI ​​model into multiple shares for the secure computation system 10, and transmits the shares corresponding to each secure computation server 100-x (S211). The secure computation system 10 performs machine learning of the AI ​​model using secure computation, and each secure computation server 100-x transmits a share of the evaluation value to the secure computation client device 200. The restoration unit 212 receives the share of the evaluation value, obtains the evaluation value of the plaintext, and acquires the corresponding performance index (S212).

[0019] The AutoML unit 220 evaluates the AI ​​model and hyperparameters according to the plaintext performance index based on the evaluation value (S221). When the learning accuracy is used as the performance index, the accuracy rate may be used as the performance index, but the precision rate, recall rate, etc. may also be used as the performance index.

[0020] Machine learning is a technology that is highly dependent on data, with optimal methods varying depending on the characteristics of the data, and from a practical standpoint, it is incompatible with secure computation, which is performed as a general process without looking at the data. The AutoML unit 220 automatically tunes, so it can repeat optimal searches even when the data is not being viewed. Furthermore, the AutoML unit 220 is controlled using plaintext, and the secure computation system 10 shares the machine learning process, achieving both efficiency and data confidentiality.

[0021] According to the secure computation client device of the present invention, AI model learning is performed using secure computation, and the pre-processing processing of the attribute selection unit 230 and the post-processing processing of the AutoML unit 220 are performed in plain text. This provides an efficient attribute selection method. In particular, the AutoML unit selects attributes to be used for learning based on performance indicators, which are plain text statistics, allowing for automatic search for the optimal machine learning model while maintaining data confidentiality. Users can select the optimal machine learning model with minimal effort, without worrying about individual learning results. Furthermore, the contents of the learning data are not leaked, except for the number of records, number of attributes, and learning model results. This allows data accumulated by multiple companies to be pooled and safely utilized, expanding the applicability of secure computation AI (such as data analysis using AI) and improving ease of implementation (such as business asset collaboration), thereby broadening the range of users. [Variation 1]

[0022] FIG. 2 also shows the processing flow of this modified example. In the first embodiment, the attribute selection unit 230 selected attributes based on their correlation with the target variable, but the possibility of factors other than correlation being considered cannot be denied. In this modified example, the attribute selection unit 230 selects a predetermined number of attributes from those excluded in the attribute selection step (S231) and adds them to the selected attributes (S232). The "predetermined number" may be, for example, two or less. The attributes may be selected randomly. By adding attributes unrelated to correlation, a wider range of attribute selections can be achieved. While the method of this modified example tends to require longer calculation times, it can achieve optimal search by adopting an intermediate behavior compared to sequential search.

[0023] [Processor, Program, Recording Medium] The functions performed by the components described herein may be implemented in circuitry or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), a CPU (a Central Processing Unit), conventional circuits, and / or combinations thereof, programmed to perform the described functions. A processor includes transistors and other circuits and is considered to be circuitry or processing circuitry. A processor may also be a programmed processor that executes a program stored in a memory.

[0024] In this specification, a circuitry, unit, or means is hardware that is programmed to realize or performs the described functions, which may be any hardware disclosed herein or any hardware known to be programmed to realize or perform the described functions.

[0025] If the hardware is a processor considered to be a type of circuitry, the circuitry, means, or unit is a combination of the hardware and software used to configure the hardware and / or processor.

[0026] The various processes described above can be implemented by loading a program that executes each step of the above method into the recording unit 2020 of the computer 2000 shown in Figure 3, and operating the control unit 2010, input unit 2030, output unit 2040, display unit 2050, etc.

[0027] The program describing the processing contents can be recorded on a computer-readable recording medium, which may be, for example, a magnetic recording device, an optical disk, a magneto-optical recording medium, a semiconductor memory, or any other suitable recording medium.

[0028] The program may be distributed by, for example, selling, transferring, lending, etc. portable recording media such as DVDs and CD-ROMs on which the program is recorded. Furthermore, the program may be stored in a storage device of a server computer, and then transferred from the server computer to other computers via a network, thereby distributing the program.

[0029] A computer that executes such a program may first temporarily store the program recorded on a portable recording medium or transferred from a server computer in its own storage device. Then, when executing a process, the computer reads the program stored on its own recording medium and executes the process in accordance with the read program. Alternatively, the computer may read the program directly from a portable recording medium and execute the process in accordance with the program. Furthermore, the computer may execute the process in accordance with the program each time a program is transferred from a server computer to the computer. Alternatively, the server computer may not transfer the program to the computer, but may instead execute the process through a so-called ASP (Application Service Provider) service, which realizes the processing function by issuing an execution instruction and obtaining the results. Furthermore, the server computer may execute the process at the terminal using a so-called SaaS (Software as a Service) service, which allows users to use part of a server computer along with the program. In this embodiment, the program includes information used for processing by an electronic computer that is equivalent to a program (such as data that is not a direct instruction to a computer but has properties that dictate computer processing).

[0030] Furthermore, in this embodiment, the device is configured by executing a predetermined program on a computer, but at least a part of the processing contents may be realized by hardware.

[0031] 10 Secure computation system 100 Secure computation server 200 Secure computation client device 210 Secure computation client unit 211 Secret sharing unit 212 Restoration unit 220 AutoML unit 230 Attribute selection unit 280 Communication unit 290 Recording unit 300 User device

Claims

1. A secure computation client device comprising: a secure computation client unit connected to a secure computation system consisting of one or more secure computation servers, which transmits shares of instruction information regarding machine learning of an AI model to the secure computation system and receives shares of evaluation values ​​of the machine learning of the AI ​​model; an attribute selection unit which selects attributes to be used in machine learning of the AI ​​model based on correlation with a target variable, sets the selected attributes as selected attributes, and instructs the secure computation client unit; and an AutoML unit which evaluates the AI ​​model and hyperparameters according to plaintext performance indicators based on the evaluation values.

2. A secure computation client device according to claim 1, wherein the attribute selection unit uses the L1 norm of the regression coefficients of the attributes in selecting the attributes.

3. A secure computation client device according to claim 1 or 2, characterized in that the attribute selection unit selects a predetermined number of attributes from the attributes excluded in the selection based on correlation with the objective variable, and adds them to the selected attributes.

4. An attribute selection method using a secure computation client device connected to a secure computation system consisting of one or more secure computation servers, comprising the steps of: selecting attributes to be used in machine learning of an AI model based on correlation with a target variable to set them as selected attributes; transmitting a share of instruction information regarding the machine learning of the AI ​​model to the secure computation system; receiving a share of an evaluation value of the machine learning of the AI ​​model; and evaluating the AI ​​model and hyperparameters according to a plaintext performance index based on the evaluation value.

Images