Secure computation device, secure computation method, and program

The secure computing device addresses scaling challenges in secure environments by aligning msb of share vectors with a reference msb, preventing overflow and underflow, achieving high-precision scaling and reducing costs, thus enhancing secure data analysis and machine learning capabilities.

WO2024241366A1PCT designated stage expired Publication Date: 2024-11-28NT T INC
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2023/018742
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-19
Publication Date
2024-11-28

AI Technical Summary

Technical Problem

Existing methods for scaling data in secure computing environments face challenges with overflow and underflow issues when using fixed-point numbers, especially when data confidentiality is maintained, as the size of data values cannot be known, leading to accuracy problems and high calculation costs with floating-point numbers.

Method used

A secure computing device with a reference msb calculation unit, msb shift unit, and scaling unit that calculates and shifts the most significant bit (msb) of share vectors to align with a reference msb, preventing overflow and underflow while keeping data secret, using msb matching to scale data efficiently within a limited range of fixed-point values.

Benefits of technology

Enables high-precision scaling without overflow or underflow, maintaining data confidentiality, and reducing calculation costs, allowing for efficient and accurate data processing in secure computing environments, expanding the scope of secure data analysis and machine learning applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023018742_28112024_PF_FP_ABST
    Figure JP2023018742_28112024_PF_FP_ABST
Patent Text Reader

Abstract

A secure computation device according to the present disclosure comprises: a reference msb calculation unit that calculates a reference msb, which is an msb serving as a reference for avoiding an overflow or underflow due to scaling; an msb shift unit that shifts the msb of all share vectors while concealing the msb so that the msb of a share vector having the largest absolute value among secretly-shared share vectors becomes equal to the reference msb; and a scaling unit that scales the share vectors having undergone the shift while keeping the share vectors concealed.
Need to check novelty before this filing date? Find Prior Art

Description

Secure computation device, secure computation method, and program

[0001] The present disclosure relates to a secure computing device, a secure computing method, and a program that scales while maintaining data confidentiality.

[0002] Conventionally, when performing data analysis, it is common to standardize or normalize data in advance. In plain text, scaling of table data can be easily performed by using data analysis libraries such as scikit-learn and R (Non-Patent Document 1).

[0003] scikit-learn developers (BSD License), “scikit-learn User Guide 6.3. Preprocessing data”, [online], [Retrieved May 15, 2023], Internet〈 URL: https: / / scikit-learn.org / stable / modules / preprocessing.html 〉

[0004] Until now, there has been no general-purpose method for scaling data with high accuracy while maintaining its confidentiality. Generally, data before scaling has a variety of values, but when data is kept confidential, the magnitude of the data values ​​is unknown. Therefore, simply scaling using the same logic as plaintext can result in unexpected underflows and overflows, reducing scaling accuracy or even making it impossible to scale correctly. While accuracy issues can sometimes be resolved by using floating-point numbers, as with common plaintext data analysis libraries, the computational cost of floating-point numbers is high in secure computation, making it desirable to use fixed-point numbers when considering processing performance.

[0005] Therefore, the object of this disclosure is to provide a secure computing device that can perform scaling without causing overflow or underflow within a limited range of values ​​that can be expressed in fixed-point numbers while maintaining the confidentiality of data.

[0006] The secure computing device of the present disclosure includes a reference msb calculation unit, an msb shift unit, and a scaling unit. The reference msb calculation unit calculates a reference msb, which is an msb that serves as a reference for avoiding overflow or underflow due to scaling. The msb shift unit shifts the msbs of all share vectors while keeping them confidential so that the msb of the share vector with the largest absolute value among the secretly shared share vectors becomes equal to the reference msb. The scaling unit scales the shifted share vector while keeping it confidential.

[0007] According to the secure computing device of the present disclosure, scaling can be performed within a limited range of values ​​that can be expressed by fixed-point numbers without causing overflow or underflow, while maintaining the confidentiality of data.

[0008] 1 is a block diagram showing the functional configuration of a secure computing apparatus according to a first embodiment. 2 is a flowchart showing the operation of the secure computing apparatus according to a first embodiment. 3 is a block diagram showing the functional configuration of a scaling unit of the secure computing apparatus according to a first embodiment. 4 is a flowchart showing the operation of the scaling unit of the secure computing apparatus according to a first embodiment. 5 is a diagram showing an example of the functional configuration of a computer.

[0009] Hereinafter, embodiments of the present disclosure will be described in detail. Note that components having the same functions are assigned the same numbers, and redundant explanations will be omitted.

[0010] The secure computing device of the first embodiment can efficiently scale data with high accuracy within a limited range of values ​​that can be expressed by fixed-point numbers without causing overflow or underflow, while keeping the data secret. There are various methods for scaling, such as standardization, normalization, and robust scaling, and the secure computing device of the first embodiment can be used for general-purpose purposes with various scaling methods.

[0011] The secure computing device of this embodiment uses a technique called MSB (most significant bit) alignment. MSB alignment is a technique in which a reference MSB, such as public information, is set, and the entire input share vector is shifted so that the MSB of the input share vector with the largest absolute value matches the reference MSB. For example, if the reference public information MSB is the 20th bit and the share vector with the largest absolute value is 14 bits, the entire input share vector is shifted by 6 bits. Since the share vector MSB and the difference between the reference MSB and the share vector MSB are kept secret, it is not possible to know the approximate bit size of the input share vector. However, the advantage is that the MSB after shifting is the same as the reference MSB, making it possible to know. By applying MSB alignment, the input share vector can be aligned to any number of bits while keeping the actual value secret, thereby preventing overflow and underflow in subsequent processing.

[0012] Also, even if you perform a process such as msb matching, where you multiply the entire vector by n, the result of subsequent scaling will not change. For example, the vector x → The whole is multiplied by n (x') → Consider the case where the following is standardized:

[0013] (x') → The average value of x → n times the average value of (x') → The variance of is x → Variance of n 2 Standardization is calculated as (data - average data value) / (standard deviation of data value), but in the numerator, both the data and the average data value are n times what they were before applying MSB adjustment, and the standard deviation of the data in the denominator is also n times (because standard deviation is the square root of variance), so the n times the numerator and denominator cancel each other out, and the standardized result is the same as if MSB adjustment had not been applied.

[0014] The functional configuration of the secure computing device 1 of this embodiment will be described below with reference to Fig. 1. As shown in the figure, the secure computing device 1 of this embodiment includes a data storage unit 10, a reference msb calculation unit 11, an msb shift unit 12, and a scaling unit 13. The data storage unit 10 is assumed to store data necessary for calculating the reference msb. The operation of each component will be described below with reference to Fig. 2.

[0015] <Reference msb Calculation Unit 11> The reference msb calculation unit 11 calculates a reference msb that is a reference msb for avoiding overflow or underflow due to scaling (S11).

[0016] <MSB Shifting Unit 12> The msb shifting unit 12 shifts the msb of all the secretly shared share vectors while keeping them secret, so that the msb of the share vector with the largest absolute value becomes equal to the reference msb (S12).

[0017] <Scaling Unit 13> The scaling unit 13 scales the shifted share vector while keeping it secret (S13). Any scaling method can be used, but for example, standardization may be used as the scaling method. Standardization is a process of setting the variance of data to 1 and the mean value to 0, and can be calculated as (data - mean value of data) / (standard deviation of data). In this case, the scaling unit 13 simply calculates (share vector - mean value of share vector) / (standard deviation of share vector).

[0018] Furthermore, for example, normalization may be used as a scaling technique. Normalization is a process of making the maximum value of data 1 and the minimum value 0, and can be calculated as (data - minimum value of data) / (maximum value of data - minimum value of data). When calculating the maximum and minimum values ​​within a vector using secure computation, the minimum and maximum values ​​can be obtained while the data remains encrypted by obtaining the first or last element after performing secure sorting. In this case, the scaling unit 13 simply calculates (share vector - minimum value of share vector) / (maximum value of share vector - minimum value of share vector).

[0019] [Variation 1] Standardization involves dividing by the standard deviation of the data, but because the standard deviation is the square root of the variance, ordinary calculations require two relatively heavy processes in secure computation: calculating the square root and dividing. However, because the computational costs of calculating the square root and the reciprocal of the square root are roughly the same, the computational cost of one division can be reduced to the cost of one multiplication by using the following calculation method: 1) calculate the variance of the data, 2) calculate the reciprocal of the square root of the variance, and 3) multiply the result of calculation 2) by (data - average value of the data) (generally, the computational cost of multiplication is smaller than the computational cost of division).

[0020] Applying the above-mentioned concept, the scaling unit 13 of the secure computing device 1 may multiply the variance of the share vector, the reciprocal of the square root of the variance, and the reciprocal of the square root by (share vector - average value of share vector). Specifically, as shown in Fig. 3, the scaling unit 13 may include a variance calculation unit 131, a square root reciprocal calculation unit 132, and a multiplication unit 133. The operation of each component will be described below with reference to Fig. 4.

[0021] <Variance Calculation Unit 131> The variance calculation unit 131 calculates the variance of the share vector (S131).

[0022] <<Square Root Reciprocal Calculation Unit 132>> The square root reciprocal calculation unit 132 calculates the reciprocal of the square root of the calculated variance (S132).

[0023] <<Multiplication Unit 133>> The multiplication unit 133 multiplies the reciprocal of the calculated square root by (share vector−average value of share vector) (S133).

[0024] <Effects of the Secure Computing Device of Example 1 and Modification 1> According to the secure computing device of Example 1 and Modification 1, scaling can be performed efficiently with high precision without causing overflow or underflow within a limited range of values ​​that can be expressed by fixed-point numbers, while keeping data secret through encryption.

[0025] Scaling is essential as a preprocessing step for various types of data analysis, such as statistical analysis and machine learning, so being able to scale data while keeping it confidential will expand the range of data analysis that can be handled with secure computing.

[0026] Furthermore, the secure computing devices of Example 1 and Modification 1 use MSB matching to enable the scaling of table data, which is difficult to process with high accuracy and high efficiency in secure computation, to be processed at high speed and with accuracy comparable to that of plaintext. According to the secure computing devices of Example 1 and Modification 1, by adjusting the number of bits after applying MSB matching, it is possible to prioritize either processing speed or accuracy as needed.

[0027] The techniques of the secure computing devices of Example 1 and Modification 1 can be applied to various scaling techniques such as standardization, normalization, and robust scaling, and can therefore be used as preprocessing for various data analyses, further expanding the scope of what can be done with data analysis while keeping data securely protected by encryption.

[0028] <Additional Notes> The device disclosed herein may, for example, be a single hardware entity that includes an input unit to which a keyboard or the like can be connected, an output unit to which an LCD display or the like can be connected, a communication unit to which a communication device (e.g., a communication cable) capable of communicating with an external device of the hardware entity can be connected, a CPU (which may also include a central processing unit, cache memory, registers, etc.), memories such as RAM and ROM, an external storage device such as a hard disk, and buses connecting these input unit, output unit, communication unit, CPU, RAM, ROM, and external storage device so as to enable data exchange. If necessary, the hardware entity may also be provided with a device (drive) capable of reading and writing recording media such as a CD-ROM. An example of a physical entity equipped with such hardware resources is a general-purpose computer.

[0029] The external storage device of the hardware entity stores the programs required to realize the above-mentioned functions and the data required for processing these programs (the programs may be stored in a ROM, which is a read-only storage device, for example, instead of an external storage device). Data obtained by processing these programs is stored in RAM, the external storage device, etc. as appropriate.

[0030] In a hardware entity, each program stored in an external storage device (or ROM, etc.) and the data required to process each program are loaded into memory as needed, and interpreted, executed, and processed by the CPU as appropriate, resulting in the CPU realizing a predetermined function (each component represented as a unit, means, etc.).

[0031] The present disclosure is not limited to the above-described embodiments, and modifications can be made as appropriate without departing from the spirit of the present disclosure. Furthermore, the processes described in the above embodiments may not only be executed in chronological order according to the order described, but may also be executed in parallel or individually depending on the processing capabilities of the devices that execute the processes or as needed.

[0032] As described above, when the processing functions of the hardware entities (apparatuses of the present disclosure) described in the above embodiments are realized by a computer, the processing contents of the functions that the hardware entities should have are described by a program, and the processing functions of the hardware entities are realized on the computer by executing this program.

[0033] The various processes described above can be implemented by loading a program that executes each step of the above method into the recording unit 10020 of the computer 10000 shown in Figure 5 and operating the control unit 10010, input unit 10030, output unit 10040, etc.

[0034] The program describing the processing contents can be recorded on a computer-readable recording medium. Examples of computer-readable recording media include magnetic recording devices, optical disks, magneto-optical recording media, and semiconductor memories. Specifically, examples of magnetic recording devices include hard disk drives, flexible disks, and magnetic tapes; optical disks include DVDs (Digital Versatile Discs), DVD-RAMs (Random Access Memory), CD-ROMs (Compact Disc Read Only Memory), and CD-Rs (Recordable / Rewritable) / RWs; magneto-optical recording media include MOs (Magneto-Optical Discs), and semiconductor memories include EEP-ROMs (Electrically Erasable and Programmable-Read Only Memory).

[0035] 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.

[0036] 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 each program transferred from the server 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. In this embodiment, the program includes information used for processing by a computer that is equivalent to a program (e.g., data that is not a direct instruction to the computer but has properties that define computer processing).

[0037] In addition, in this embodiment, a hardware entity is configured by executing a predetermined program on a computer, but at least a part of the processing contents may be realized by hardware.

Claims

1. A secure computing device comprising: a reference msb calculation unit that calculates a reference msb, which is a reference msb for avoiding overflow or underflow due to scaling; an msb shift unit that shifts the msbs of all share vectors while keeping them secret, so that the msb of the share vector with the largest absolute value among the secretly shared share vectors becomes equal to the reference msb; and a scaling unit that scales the shifted share vector while keeping it secret.

2. A secure computing device according to claim 1, wherein the scaling is standardization, and the scaling unit calculates (share vector-average value of share vectors) / (standard deviation of share vectors).

3. A secure computing device as described in claim 1, wherein the scaling is standardization, and the scaling unit includes: a variance calculation unit that calculates the variance of the share vector; a square root reciprocal calculation unit that calculates the reciprocal of the square root of the calculated variance; and a multiplication unit that multiplies the reciprocal of the calculated square root by (share vector - average value of share vector).

4. A secure computing device according to claim 1, wherein the scaling is normalization, and the scaling unit calculates (share vector-minimum value of share vector) / (maximum value of share vector-minimum value of share vector).

5. A secure computation method executed by a secure computing device, comprising: a step of calculating a reference msb, which is a standard msb for avoiding overflow or underflow due to scaling; a step of shifting the msb of all share vectors while keeping it secret, so that the msb of the share vector with the largest absolute value among the secretly shared share vectors becomes equal to the reference msb; and a step of scaling the shifted share vector while keeping it secret.

6. A program for causing a computer to function as a secure computing device according to any one of claims 1 to 4.

Citation Information

Patent Citations

  • Hidden sigmoid function calculation system, hidden logistic regression calculation system, hidden sigmoid function calculation device, hidden logistic regression calculation device, hidden sigmoid function calculation method, hidden logistic regression calculation method, and program

    WO2020071441A1

  • Integrated analysis method, integrated analysis device, and integrated analysis program

    WO2021090789A1

  • Secure computation device, secure computation method, and program

    WO2021149104A1

  • Confidential MSB normalization system, distributed processing device, confidential MSB normalization method, and program

    WO2022079891A1