Secret statistical processing systems, methods, statistical processing apparatus and program

Patent metadata

Field	Value
Publication number	US-11880489-B2
Application number	US-201917274187-A
Country	US
Kind code	B2
Filing date	Aug 28, 2019
Priority date	Sep 10, 2018
Publication date	Jan 23, 2024
Grant date	Jan 23, 2024

Abstract

Provided is a technique for performing statistical processing such as processing for obtaining parameters of logistic regression analysis faster than before. A secure statistical processing system includes a cross tabulation table computing device 2 that performs secure computation on a cross tabulation table in which frequencies are in plain texts while keeping each record concealed; and a statistical processing device 3 that performs predetermined statistical processing using the cross tabulation table in which frequencies are in plain texts. The cross tabulation table computing device 2 may include a plurality of secure computation devices 221, . . . , 22 N that perform secure computation on a cross tabulation table in which frequencies are fragments subjected to secret sharing while keeping each record concealed, and a management device 21 that restores the fragments to compute the cross tabulation table in which frequencies are in plain texts.

First claim

The invention claimed is:

1. A secure statistical processing system, comprising:
cross tabulation table computing circuitry configured to perform secure computation on a cross tabulation table in which each record is concealed and frequencies of each record having a predetermined value are in plain text; and
statistical processing circuitry configured to perform predetermined statistical processing using the cross tabulation table in which frequencies are in plain texts,
wherein the statistical processing circuitry performs processing of calculating parameters of logistic regression analysis, and
the cross tabulation table computing circuitry includes:
a plurality of secure computation circuitries that performs secure computation on a cross tabulation table in which frequencies are fragments subjected to secret sharing while keeping each record concealed, and
management circuitry that restores the fragments to compute the cross tabulation table in which frequencies are in plain texts, wherein
the plurality of secure computation circuitries randomly exchange records in performing the secure computation,
the plurality of secure computation circuitries perform the secure computation in parallel, and
the plurality of computation circuitries calculate the parameters of the logistic regression analysis by maximizing an error function whose arguments are the parameters.

2. The statistical processing circuitry of the secure statistical processing system according to claim 1 .

3. A secure statistical processing method, comprising:
by cross tabulation table computing circuitry, performing secure computation on a cross tabulation table in which each record is concealed and frequencies of each record having a predetermined value are in plain text; and
by statistical processing circuitry, performing predetermined statistical processing using the cross tabulation table in which frequencies are in plain texts, including calculating parameters of logistic regression analysis, wherein
the method further includes:
performing, by a plurality of secure computation circuitries, secure computation on a cross tabulation table in which frequencies are fragments subjected to secret sharing while keeping each record concealed, and
restoring the fragments to compute the cross tabulation table in which frequencies are in plain texts, wherein
the plurality of secure computation circuitries randomly exchange records in performing the secure computation,
the plurality of secure computation circuitries perform the secure computation in parallel, and
the plurality of computation circuitries calculate the parameters of the logistic regression analysis by maximizing an error function whose arguments are the parameters.

4. A non-transitory computer-readable medium storing a program, that, when executed by a computer, causes the computer to perform a method according to claim 3 .

5. The statistical processing system according to claim 1 , wherein the predetermined statistical processing performed by the processing circuitry is unsecured statistical processing.

6. The statistical processing system according to claim 5 , wherein the records include health insurance information.

7. The statistical processing system according to claim 6 , wherein the health insurance information includes height, weight, and blood pressure information.

8. The statistical processing system according to claim 7 , wherein the predetermined statistical processing included divisions and exponential functions.

9. The statistical processing system according to claim 1 , wherein in calculation the parameters of the logistic regression analysis by maximizing the error function, the plurality of computation circuitries partially differentiate the error function.

10. The statistical processing system according to claim 9 , wherein the plurality of computation circuitries partially differentiate the error function with respect to the parameters to generate a set of partial derivative functions.

11. The statistical processing system according to claim 10 , wherein the plurality of computation circuitries calculate the parameters such that the set of partial derivative functions are substantially zero.

Assignees

• Nippon Telegraph & Telephone

Inventors

• Tanaka Satoshi
• Miyajima Asami
• Takahashi Gen
• Ikarashi Dai
• Hamada Koki
• Kikuchi Ryo
• Fuji Hitoshi

Classifications

• G06F21/6245 Primary

  Protecting personal data, e.g. for financial or medical purposes · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G09C1/00

  Apparatus or methods whereby a given sequence of signs, e.g. an intelligible text, is transformed into an unintelligible sequence of signs by transposing the signs or groups of signs or by replacing them by others according to a predetermined system (cryptographic typewriters G09C3/00) · CPC title

• H04L2209/46

  Secure multiparty computation, e.g. millionaire problem · CPC title