CONFIDENTIAL SORT SYSTEM AND METHOD

Abstract

The present invention provides a technique for performing confidential sort at a faster speed than in the prior art. A confidential sort system comprises first to Mth apparatuses. The first to Mth apparatuses obtain inverse substitution [[σ0−1]] of L-bit stable sort of {→k0}. The first to Mth apparatuses perform, on i=1, . . . , N−1, a process of converting [[σi−1−1]] to hybrid substitution to obtain {σi−1−1}, a process of inversely substituting {→ki} with {σi−1−1} to obtain {σi−1→ki}, a process of obtaining inverse substitution [[σ′i−1]] of L-bit stable sort of [[σi−1→ki]], a process of synthesizing {σi−1−1} with [[σ′i−1]] to obtain [[σi−1]]:=[[σi−1−1σ′i−1]], and a process of converting [[σN−1−1]] to hybrid substitution to obtain {σN−1−1}. The first to Mth apparatuses inversely substitute [[→v]] with {σN−1−1} and output [[σN−1→v]].

Description

TECHNICAL FIELD

The present invention relates to an information security technique.

BACKGROUND ART

As a conventional confidential sort technique, a technique described in Non-patent literature 1 is known.

PRIOR ART LITERATURE

Non-Patent Literature

Non-patent literature 1: Dai Ikarashi, Ryo Kikuchi, Koki Hamada, Koji Chida, “An Unconditionally Private and Correct MPC Construction against the Active Adversary on Multiple Fields and an Application to Fast Secure Sorting,” In SCIS2015, 2015.

SUMMARY OF THE INVENTION

Problems to be Solved by the Invention

The technique described in Non-patent literature 1 above had a large communication amount associated with sort processing, and so sort processing could be slowed down with communication becoming a bottleneck.

It is an object of the present invention to provide a confidential sort system and method capable of performing a confidential sort at a speed faster than in the prior art.

Means to Solve the Problems

A confidential sort system according to an aspect of the present invention comprises first, second, . . . , and Mth apparatuses. Assuming that a body of data to be sorted is a value ^→v, a tag that determines order after the sort is a key, a bit length of the key is L′, N is a predetermined positive integer, an (n+1)th (n=0, . . . , N−1) bit string when the key is divided into N bit strings is ^→k_n, an arbitrary value or substitution is x^A, {x^A} is a replicated secret variance of x^A, an arbitrary value is x^A, [[x^A]] is a secret variance having homomorphism of x^A and M is a predetermined positive integer of 2 or more, the first, second, . . . , and Mth apparatuses obtain inverse substitution [[σ₀⁻¹]] of L-bit stable sort of {^→k₀}. The first, second, . . . , and Mth apparatuses perform, on i=1, . . . , N−1, a process of obtaining {σ_i−1⁻¹} by converting [[δ_i−1⁻¹]] to hybrid substitution, a process of obtaining {σ_i−1^→k_i} by inversely substituting {^→k_i} using {σ_i−1⁻¹}, a process of obtaining inverse substitution [[σ′_i⁻¹]] of L-bit stable sort of [[σ_i−1^→k_i]], a process of obtaining [[σ_i⁻¹]]:=[[(σ_i−1⁻¹σ′_i⁻¹]] by synthesizing {σ_i−1⁻¹} with [[σ′_i−1⁻¹]], and a process of obtaining {σ_N−1⁻¹} by converting [[σ_i−1⁻¹]] to hybrid substitution. The first, second, . . . , and Mth apparatuses output [[σ_N−1^→v]] by inversely substituting [[^→v]] using {σ_N−1⁻¹}.

A confidential sort system according to an aspect of the present invention comprises first, second, . . . , and Mth apparatuses. Assuming that a body of data to be sorted is value ^→v, a tag that determines order after the sort is a key, a bit length of the key is L′, N is a predetermined positive integer, an (n+1)th (n=0, . . . , N−1) bit string when the key is divided into N bit strings is →k_n, an arbitrary value or substitution is x^A, {x^A} is a replicated secret variance of x^A, an arbitrary value is x^A, [[x^A]] is a secret variance having homomorphism of x^A, <x^A> is a semi-public value, [x^A] is a (2,2) additive secret variance, {^→k_i}^{(Z_2){circumflex over ( )}L}=({^→k_i,0}^Z_2, {^→k_i,1}^Z_2, . . . , {^→k_{i, L−1}}^Z_2) (i=0, . . . , N−1), q is a predetermined positive integer, an arbitrary vector is ^→f^A, (^→f^A), is a uth element of vector ^→f^A, s₀:=0, s_j:=Σ_0≤u<m(^→f_j−1)_u+s_j−1 for j=1, . . . , 3, the number of elements of vector ^→f_j−1 is m, 01 represents a set of the first and second apparatuses, 12 represents a set of the second and third apparatuses, 20 represents a set of the third and first apparatuses, G is an arbitrary group, ring or body, P is a set of apparatuses, [x^A]^G,P represents a share of [x^A] represented on G of P, π is predetermined substitution, {π}_P means a sub-share of {π} shared by P, {π}^01,12 represents a replicated secret variance in which π={π}₁₂{π}₀₁, <πσ_i⁻¹>²⁰ represents that πσ_i⁻¹ is shared by the third and first apparatuses, the first, second and third apparatuses perform a process of converting, through mod 2→mod q conversion, {^→k_0,0}^Z_2, {^→k_0,1}^Z_2, . . . , {^→k_{0, L−1}}^Z_2 to [[^→k_0,0]]^Z_q, [[^→k_0,1]]^Z_q, . . . , [[^→k_{0, L−1}]]^Z_q, a process of calculating [[K_D]]:=[[Π_w∈D^→k_0,w]]^Z_q for each set of a positive integer η satisfying η<┌log L┐ and D satisfying D⊆Z_L and 2^η+1≤|D|≤min(2^η+1, L), a process of calculating [[^→f_j]]^Z_q:=Π_j<L[[^→k′_0,w]] for each j where j<2^L assuming the wth bit of bit expression of j as j_w, and assuming ^→k′_0,w to be ^→k′_0,w=1−^→k_0,w when j_w=0 and ^→k′_0,w=^→k_0,w when j_w=1, a process of calculating [[(^→f′_j)_u]]^Z_q:=Σ_0≤t<u[[(^→f_j)_t]]^Z_q+[[(s_j)]]^Z_q, a process of calculating [σ₀⁻¹]^Z_q,01:=Σ_{0≤j<2{circumflex over ( )}L}[^→f′_j^→f_j]^Z_q,01 through a (2,2) output product sum operation, and a process of calculating <πσ₀⁻¹>²⁰:={π}^01,12[σ₀⁻¹]^Z_q,01. The first, second and third apparatuses perform a process of converting {^→k_i,0}^Z_2, {^→k_i,1}^Z_2, . . . , {^→k_{i, L−1}}^Z_2 to (2,2) additive secret variance [^→k_i,0]^Z_2,01, [^→k_i,1]^Z_2,01, . . . , [^→k_{i, L−1}]^Z_2,01 assuming i=1, . . . , N−1, a process of inversely applying{σ_i−1⁻¹} to [^→k_i,0]^Z_2,01, [^→k_i,1]^Z_2,01 . . . , [^→k_{i, L−1}]^Z_2,01 to obtain [^→b₀]^Z_2,20:=[σ_i−1k_i,0]^Z_2,20, [^→b₁]^Z_2,20:=[σ_i−1k_i,1]^Z_2,20, . . . , [^→b_L−1]^Z_2,20:=[σ_i−1k_{i, L−1}]^Z_2,20, a process of converting [^→b₀]^Z_2,20, [^→b₁]^Z_2,20, . . . , [^→b_L−1]^Z_2,20 to [[^→b₀]]^Z_q, [[^→b₁]]^Z_q, . . . , [^→b_L−1]^Z_q through mod 2→mod q conversion, a process of calculating [[K_D]]:=[[Π_w∈D^→k_0,w]]^Z_q for each set of a positive integer η satisfying η<┌log L┐ and D⊆Z_L and D satisfying 2^η+1≤|D|≤min(2^η+1, L), a process of calculating [[^→f_j]]^Z_q:=Π_j<L[[^→k′_i,w]] for each j where j<2^L assuming the wth bit of bit expression of j as j_w, and assuming as when j_w=1 and ^→k′_i,w=^→k_i,w when j_w=1, a process of calculating [[(^→f′_j)_u)]]^Z_q:=Σ_0≤t≤u[[(^→f_j)_t]]^Z_q+[[s_j]]^Z_q, a process of calculating [σ′⁻¹]^Z_q,20:=Σ_{0≤j<2{circumflex over ( )}L}[^→f′_j^→f_j]^Z_q,20 through a (2,2) output product sum operation and a process of calculating [σ⁻¹]^Z_q,01:=[σ_i−1⁻¹σ′⁻¹]^Z_q,01 by applying {σ_i−1⁻¹} to [σ⁻¹]^Z_q,20 and a process of calculating <πσ_i⁻¹>²⁰:={π}^01,12[σ⁻¹]^Z_q,01 on i=1, . . . , N−1. The first, second, . . . , and Mth apparatuses inversely substitute [[^→v]] as {σ⁻¹}=(<πσ_N−1⁻¹>²⁰, {π}^01,12) with {σ⁻¹} and output [[σ^→v]].

Effects of the Invention

It is possible to perform a confidential sort at a speed faster than in the prior art.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram illustrating an example of a confidential sort system of a first embodiment;

FIG. 2 is a diagram for describing the first embodiment;

FIG. 3 is a diagram for describing the first embodiment;

FIG. 4 is a diagram for describing the first embodiment;

FIG. 5 is a diagram for describing the first embodiment;

FIG. 6 is a block diagram illustrating an example of a confidential sort system according to a second embodiment;

FIG. 7 is a diagram for describing the second embodiment;

FIG. 8 is a diagram for describing the second embodiment;

FIG. 9 is a diagram for describing the second embodiment;

FIG. 10 is a diagram for describing the second embodiment;

FIG. 11 is a diagram for describing the second embodiment;

FIG. 12 is a diagram for describing the second embodiment;

FIG. 13 is a diagram for describing the second embodiment;

FIG. 14 is a diagram for describing the second embodiment;

FIG. 15 is a diagram for describing the second embodiment;

FIG. 16 is a diagram for describing the second embodiment;

FIG. 17 is a diagram for describing the second embodiment; and

FIG. 18 is a diagram for describing the second embodiment.

DETAILED DESCRIPTION OF THE EMBODIMENTS

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings. Note that in mathematical expressions or the like, when the base of a log is omitted, the base is assumed to be 2.

[Preparations]

Examples of sort targets include a key and a value. The key is a tag to determine order after a sort and the value is the body of data to be sorted. The key and the value may be identical.

A secret variance refers to a set of virtually collected secretly distributed shares of all parties.

A result of applying substitution π to vector →x is multiplicatively written as π^→x. Furthermore, the integer vector is also handled as substitution and multiplicative notation ^→y^→x of a vector represents substitution of ^→x by ^→y. Note that for simplification of description, “^→” that means a vector may be omitted.

Symbols are defined as follows.

m: The number of elements of a vector to be sorted and substituted.

L′: Bit length of key.

p, q: Prime number.

|p|, |q|: Bit length of prime number.

[[x]]: Secret variance having homomorphism. That is, when an arbitrary value is assumed as x^A, [[x^A]] is a secret variance having homomorphism of x^A.

{x}: Notation explicitly showing that this is a replicated secret variance. That is, when an arbitrary value or substitution is assumed as x^A, {x^A} is a replicated secret variance of x^A.

[x]: (2,2) additive secret variance.

<x>: Semi-public value. That is, plain text shared by k parties.

[[X]]: Set of secret variances on X.

[x]^X,P: First subscript on the shoulder of the above-described variance represents a group/ring/body and the second subscript represents a party set having shares. That is, assuming that G is an arbitrary group, ring or body and P is a set of apparatuses, [x]^G,P represents a share of [x] owned by P and represented on G.

^→x, [[^→x]]^X: Vector having a length of m and a secret variance thereof.

P′: Set of whole party.

01,12,20: When used as subscripts, these numbers represent a party set of parties 0 and 1, parties 1 and 2 and parties 2 and 0. For example, 01 represents a set of the first and second apparatuses, 12 represents a set of the second and third apparatuses and 20 represents a set of the third and first apparatuses. Note that the first apparatus may be represented by P₀, the second apparatus may be represented by P₁ and the third apparatus may be represented by P₂.

[[x]]_P: Share of party P.

{π}: Replicated secret variance of substitution π.

{π}_P: Fraction (subshare) of {π} shared by P∈P′.

Replicated secret variance of substitution with subscript such as {π}^01,12,20:

Replicated secret variance which becomes π={π}₂₀{π}₁₂{π}₀₁. That is, notation when even the application order of substitution is taken into consideration. Replicated secret variance of substitution with fewer than three subscripts such as {π}^01,12: Replicated secret variance which becomes π={π}₁₂{π}₀₁.

Here, the replicated secret variance is a secret variance sharing the same value within a party set for each of a plurality of party sets. For example, assuming that (2,3)-replicated secret variance is a=a₀₁+a₁₂+a₂₀, shares of the respective parties are (a₂₀,a₀₁), (a₀₁,a₁₂) and (a₁₂,a₂₀). Each a₀₁, a₁₂ or a₂₀ is called “subshare.” Here, using the fact that the substitution is a group, the substitution is also used as a secret variance to be kept confidential.

A sort is a kind of substitution. Here, a theory of substitution in secret calculation will be configured and an efficient and simple cardinal number sort will be configured based on the theory.

Substitution has never been handled systematically in secret calculation. Here, substitution will be systematically adjusted so that substitution may be handled more freely. First of all, substitution itself and substitution in existing secret calculation will be reviewed and the theory of substitution will be developed based thereon.

“Substitution” is a mathematical structure that expresses rearrangement. For example, substitution (0,2,1) rearranges vector (10,5,2) to (10,2,5).

Substitution is known to form a non-exchangeable group. That is, a unit element exists (identical substitution, substitution without changing arrangement), an inverse element exists (after substitution, substitution with inverse element returns to the original arrangement) and an associative law holds.

Random substitution in secret calculation has been proposed and improved independently by the author et al. and Laur (e.g., see Reference Documents 1 and 2). Since the replicated secret variance can be configured of a group, it is possible to consider a replicated secret variance of substitution, and in the case of, for example, (2,3)-replicated secret variance, when π=π₂₀π₁₂π₀₁, respective variances are (π₂₀,π₀₁), (π₀₁; π₁₂) and (π₁₂,π₂₀). All the above random substitutions can be regarded as protocols that apply this substitution.

• [Reference Document 1]
• Koki Hamada, Dai Ikarashi, Koji Chida, Katsumi Takahashi, “A Random Permutation Protocol on Three-Party Secure Function Evaluation,” CSS2010 (2010).
• [Reference Document 2]
• Laur, S., Willemson, J. and Zhang, B., “Round-Efficient Oblivious Database Manipulation”, ISC (Lai, X., Zhou, J. and Li, H., eds.), Lecture Notes in Computer Science, Vol. 7001, Springer, pp. 262-277 (2011).

When inverse elements ({π}⁰¹)⁻¹, ({π}¹²)⁻¹, ({π}²⁰)⁻¹ of the respective subshares {π}⁰¹, {π}¹², {π}²⁰ of substitution {π} are executed so as to be applied in reverse order according to a random substitution protocol, the same effect as that when substitution is performed with inverse matrix {π}⁻¹ (e.g., see Reference Document 3). This can be used for processing such as performing random substitution once and returning to the original arrangement.

• [Reference Document 3]
• Naoto Kiribuchi, Dai Ikarashi, Gembu Morohashi, Koki Hamada, “An Efficient Equi-join Algorithm for Secure Computation and Its Implementation toward Secure Comprehensive Analyses of Users' Attribute and History Information,” CSS2016 (2016).

A replicated secret variance of substitution is called “native substitution” in the present invention. The native substitution alone can only express random substitution but cannot handle a wider variety of substitution. Therefore, the present specification will introduce two types of substitution: index substitution and hybrid substitution.

A vector with an integer variance having length in and including different values of 0 to m−1 as elements is called “index substitution.” Although index substitution cannot be directly applied as substitution, it is possible to synthesize substitution with native substitution by applying native substitution. When, for example, application of {ρ} to [[π]] results in [[ρπ]], and this gives synthesized index substitution. When synthesis is applied to index substitution [[I]] of identical substitution I, it is also possible to convert native substitution {π} to index substitution [[π]].

Hybrid substitution is a set of native substitution and substitution of plain text. It is assumed that ({ρ},ρ⁻¹π) is written as {{π}}, and this is called “hybrid substitution.” Since conversion from hybrid substitution to native substitution is an off-line process, {{π}} may be written as {π}. Since the hybrid substitution is a set of native substitution and substitution of plain text, when hybrid substitution is applied in order, application and inverse application are made possible. Furthermore, since application is possible, it is possible to synthesize index substitution with substitution and perform conversion to index substitution.

With hybrid substitution, conversion from index substitution is further possible. Applying {ρ⁻¹} to [[π]] results in [[ρ⁻¹π]] and obtains ρ⁻¹π by public disclosure. ({ρ},ρ⁻¹π) constitutes hybrid substitution.

Since index substitution can handle any substitution other than random substitution, hybrid substitution can also keep confidential non-random substitutions. Hybrid substitution is necessary to apply non-random substitution with confidentiality.

Hybrid substitution can also be converted to native substitution. When {ρ} is written as {ρ}^01,12,20, ρ⁻¹π may be synthesized with {ρ}₂₀^01,12,20. Since ρ⁻¹π is a public value, this process is off line.

Note that since conversion from hybrid substitution to native substitution is an off-line process, according to a protocol with abstract granularity, hybrid substitution is identified with native substitution and written as {π}.

The following summarizes the above-described simply configurable substitution operations. It is possible to freely perform four operations of substitution, inverse substitution, conversion and synthesis not only via native substitution but also via index substitution and hybrid substitution. It should be noted that synthesis requires index substitution and substitution/inverse substitution after the synthesis requires hybrid substitution.

[[Substitution Application]]

1. Native substitution: {π}[[x]]=[[x]] (random substitution protocol)

2. Index substitution: performed not directly but via hybrid substitution.

3. Hybrid substitution: ({ρ}ρ⁻¹π[[x]])

[[Inverse Substitution Application]]

1. Native substitution: {π}[[x]]

2. Index substitution: performed not directly but via hybrid substitution.

3. Hybrid substitution: (ρ⁻¹)⁻¹{ρ}⁻¹[[x]]

[[Conversion]]

1. Native substitution →index substitution: {π}[[I]]=[[π]]

2. Index substitution →native substitution: via hybrid substitution.

3. Native substitution →hybrid substitution: ({π},I)

4. Hybrid substitution →native substitution: synthesize ρ⁻¹ π with {ρ}₂₀^01,12,20 of {ρ}^01,12,20.

5. Index substitution→hybrid substitution: {{π}}[[I]]={{π}}

6. Hybrid substitution →index substitution: {{π}}[[I]]=[[π]]

[[Synthesis]]

1. Index substitution and native substitution: {{ρ}}[[π]]=[[ρπ]]

2. Index substitution and hybrid substitution: {{ρ}}[[π]]

3. Other combinations are performed via index substitution.

First Embodiment

As shown in FIG. 1, a confidential sort system according to a first embodiment is provided with, for example, a first apparatus 1, a second apparatus 2, . . . , and an Mth apparatus M. M is a predetermined positive integer of 2 or more. In this example, there are M parties and the M parties are the first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M.

A confidential sort method of the first embodiment is implemented, for example, by the first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M executing processes under Scheme 4-1 in FIG. 2 and Scheme 4-2 in FIG. 3. Hereinafter, when there is a mention “the first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M perform ◯◯,” this means that the first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M jointly perform ◯◯ through secret calculation. Of course, when a process of ◯◯ does not require secret calculation, the first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M need not perform secret calculation.

The first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M perform a process under Scheme 4-1 shown in FIG. 2 and thereby perform a confidential sort process on a key first.

In “1:” under Scheme 4-1, the first, second, . . . , and Mth apparatuses obtain inverse substitution [[σ₀⁻¹]] of L-bit stable sort of {^→k₀} (“1:” under Scheme 4-1). Here, it is assumed that a bit length of a key is L′, L is a predetermined positive integer, N is a positive integer satisfying NL=L′ and an (n+1)th (n=0, . . . , N−1) bit string when the key is divided into bit strings having a bit length L is ^→k_n. Inverse substitution of the L-bit stable sort can be performed under Scheme 3-1 in FIG. 4 and Scheme 3-2 in FIG. 5. In FIG. 4 and FIG. 5, [[k]]^{(Z_2){circumflex over ( )}L}=([[k₀]]^Z_2, [[k₁]]^Z_2, . . . , [[K_L−1]]^Z_2) and [[f]]^{(Z_q){circumflex over ( )}(2{circumflex over ( )}L)}=([[f₀]]^Z_q, [[f₁]]^Z_q, . . . , [[f_{(2{circumflex over ( )}L)−1)}]]^Z_q). “(f_i)_i^” represents an ith element of a vector j. Reference character m denotes the number of elements of a vector to be sorted and substituted as described above. More specifically, m in FIG. 4 is the number of elements of k_j and m in FIG. 5 is the number of elements of f_j.

In FIG. 4, mod 2→mod q conversion can be performed, for example, under Scheme 3-3 in FIG. 13. “reveal( )” in FIG. 13 represents reconstructing and publicly disclosing a secret variance in parentheses of reveal.

In “2:” to [6:] under Scheme 4-1, the first, second, . . . , and Mth apparatuses perform processes from “3:” to “6:” under Scheme 4-1 on each i=1, . . . , N.

That is, the first, second, . . . , and Mth apparatuses convert [[σ_i−1⁻¹]] to hybrid substitution to obtain {σ_i−1⁻¹} assuming i=1, . . . , N−1 (“3:” under Scheme 4-1).

The first, second, . . . , and Mth apparatuses inversely substitute {^→k_i} with {σ_i−1⁻¹} to obtain {σ_i−1^→k_i} assuming i=1, . . . , N−1 (“4:” under Scheme 4-1).

The first, second, . . . , and Mth apparatuses obtain inverse substitution [[σ′_i⁻¹]] of the L-bit stable sort of [[σ_i−1^→k_i]] assuming i=1, . . . , N−1 (“5:” under Scheme 4-1).

The first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M synthesize {σ_i−1⁻¹} with [[σ′_i⁻¹]] to obtain [[σ_i⁻¹]]:=[[σ_i−1⁻¹σ′_i⁻¹]] assuming i=1, . . . , N−1 (“6:” under Scheme 4-1).

In “7:” under Scheme 4-1, the first, second, . . . , and Mth apparatuses convert [[σ_N−1⁻¹]] to hybrid substitution to obtain {σ_N−1⁻¹} (“7:” under Scheme 4-1).

Next, the first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M perform a process under Scheme 4-2 shown in FIG. 3 and thereby perform a confidential sort process on a value.

That is, the first apparatus 1, the second apparatus 2, . . . , and the Mth apparatus M inversely substitute [[^→v]] with {σ_N−1⁻¹} in “1:” under Scheme 4-2 shown in FIG. 3 and output [[σ_N−1^→v]] (“1:” under Scheme 4-2). “^→v” is the value which is the body of data to be sorted.

In this way, using inverse substitution, it is possible to simplify the algorithm of confidential sort compared to the prior art. This allows a confidential sort to be performed faster than in the prior art.

Note that when the secret calculation of the first embodiment is performed in a so-called passive (security at which an attacker may peep but not perform any illegal process) version, it is possible to use a Shamir secret variance or replicated secret variance.

When a Shamir secret variance is used, the method described in Reference Document 4 (e.g., 2.5, Shuffling Protocol) may be used to perform random substitution and inverse substitution. The method described in Reference Document 5 may be used to perform public disclosure. Public value output random substitution may be publicly disclosed using the method described in Reference Document 5 after performing random substitution using the method described in Reference Document 4. Addition, multiplication and product sum operations may be performed using the method described in Reference Document 6 (e.g., The Computation Stage). An L-bit stable sort may be performed using the method described in Reference Document 2 (e.g., Schemes 6+7).

When a replicated secret variance is used, random substitution and inverse substitution may be performed using the method described in Reference Document 4 (e.g., 2.5, Shuffling Protocol). Public disclosure may be performed using the method described in Reference Document 7 (e.g., cramer 2005, 2. Preliminaries, 2. Replicated Secret-sharing). Public value output random substitution may be publicly disclosed using the method described in Reference Document 7 after performing random substitution using the method described in Reference Document 4. Addition, multiplication and product sum operations may be performed using the method described in Reference Document 8 (e.g., protocols 1 and 7). An L-bit stable sort may be performed using the method described in Reference Document 2 (e.g., Schemes 6+7).

When performing secret calculation of the first embodiment using a so-called active (safe even when an attacker performs an illegal process) version, a Shamir secret variance can be used.

In the case of performing secret calculation of the first embodiment in the active version, if a Shamir secret variance is used, random substitution and inverse substitution may be performed using the method described in Reference Document 9 (e.g., Scheme 6). Public disclosure may be performed using the method described in Reference Document 10 (e.g., Scheme 12). Public value output random substitution may be publicly disclosed using the method described in Reference Document 10 after performing random substitution using the method described in Reference Document 9. Addition, multiplication and product sum operations may be performed using the method described in Reference Document 9 (e.g., Schemes 3 to 5). An L-bit stable sort may be performed using the method described in Reference Document 2 (e.g., Schemes 6+7).

• [Reference Document 4] Koki Hamada, Ryo Kikuchi, Dai Ikarashi, Koji Chida, Katsumi Takahashi, “Practically Efficient Multi-party Sorting Protocols from Comparison Sort Algorithms”, ICISC 2012: 202-216
• [Reference Document 5] Adi Shamir, “How to Share a Secret”, Commun. ACM 22(11): 612-613 (1979)
• [Reference Document 6] Michael Ben-Or, Shafi Goldwasser, Avi Wigderson, “Completeness Theorems for Non-Cryptographic Fault-Tolerant Distributed Computation (Extended Abstract)”, STOC 1988: 1-10
• [Reference Document 7] Ronald Cramer, Ivan Damgard, Yuval Ishai, “Share Conversion, Pseudorandom Secret-Sharing and Applications to Secure Computation”, TCC 2005: 342-362
• [Reference Document 8] Dai Ikarashi, Koji Chida, Koki Hamada, Katsumi Takahashi, “Secure Database Operations Using An Improved 3-party Verifiable Secure Function Evaluation.”
• [Reference Document 9] Dai Ikarashi, Ryo Kikuchi, Koki Hamada, Koji Chida, “An Unconditionally Private and Correct MPC Construction against the Active Adversary on Multiple Fields and an Application to Fast Secure Sorting”, In SCIS2015, 2015.
• [Reference Document 10] D. Ikarashi, R. Kikuchi, K. Hamada, and K. Chida, “Actively private and correct MPC scheme in t<n/2 from passively secure schemes with small overhead”, IACR Cryptology ePrint Archive, 2014: 304, 2014.

Second Embodiment

As shown in FIG. 6, a second embodiment is provided with, for example, a first apparatus 1, a second apparatus 2 and a third apparatus 3. In this example, there are three parties, and the three parties are the first apparatus 1, the second apparatus 2 and the third apparatus M respectively.

A confidential sort method according to the second embodiment is implemented, for example, by the first apparatus 1, the second apparatus 2 and the third apparatus 3 performing respective processes under Scheme 5 in FIG. 7 and Scheme 5-3 in FIG. 8. Hereinafter, when there is a mention “the first apparatus 1, the second apparatus 2 and the third apparatus 3 perform ◯◯,” this means that the first apparatus 1, the second apparatus 2 and the third apparatus 3 jointly perform ◯◯ through secret calculation. Of course, when the process of ◯◯ does not require secret calculation, the first apparatus 1, the second apparatus 2 and the third apparatus 3 need not perform secret calculation.

First, the first apparatus 1, the second apparatus 2 and the third apparatus 3 perform a process under Scheme 5 shown in FIG. 7 and thereby perform a confidential sort process on a key first.

In “1:” under Scheme 5, the first apparatus, the second apparatus and the third apparatus 3 perform the process under Scheme 5-1 on {^→k₀}^{(Z_2){circumflex over ( )}L} and thereby obtain {σ₀⁻¹}=(<πσ₀⁻¹>²⁰, {π}^01,12) (“1:” under Scheme 5).

Here, assuming that i=0, . . . , N−1, {^→k_i}^{(Z_2){circumflex over ( )}L}=({^→k_i,0}^Z_2, {^→k_i,1}^Z_2, . . . {^→k_{i, L−1}}^Z_2). It is also assumed that a bit length of a key is L′, L is a predetermined positive integer, N is a positive integer satisfying NL=L′, and an (n+1)th (n=0, . . . , N−1) bit string when the key is divided into bit strings having a bit length L is ^→k_n.

In “2:” and “3:” under Scheme 5, the first apparatus, the second apparatus and the third apparatus 3 perform a process under Scheme 5-2 on {^→k_i}^{(Z_2){circumflex over ( )}L} for i=0, . . . , N−1 to finally obtain {σ_N−1⁻¹}=(<πσ_N−1⁻¹>²⁰, {π}^01,12) (“2:” and “3:” under Scheme 5).

Next, the first apparatus 1, the second apparatus 2 and the third apparatus 3 perform a process under Scheme 5-3 shown in FIG. 8 and thereby perform a confidential sort process on a value.

That is, the first apparatus 1, the second apparatus 2 and the third apparatus 3 inversely substitute [[^→v]] with {σ_N−1⁻¹} in “1:” under Scheme 5-3 shown in FIG. 8 and output [[σ_N−1^→v]] (“1:” under Scheme 5-3). “^→v” is the value which is the body of data to be sorted.

Hereinafter, Scheme 5-1 in FIG. 9 and FIG. 10 will be described.

In “1:” under Scheme 5-1, the first apparatus, the second apparatus and the third apparatus 3 convert {^→k_0,0}^Z_2, {^→k_0,1}^Z_2, . . . , {^→k_{0, L−1}}^Z_2 to [[^→k_0,0]]^Z_q, [[^→k_0,1]]^Z_q, . . . , [[^→k_{0, L−1}]]^Z_q through mod 2 mod q conversion (“1:” under Scheme 5). Here, q is assumed to be a predetermined positive integer. A mod 2→mod q conversion can be performed under Scheme 3-3 in FIG. 13.

In “2:” to “5:” under Scheme 5-1, the first apparatus, the second apparatus and the third apparatus 3 calculate [[K_D]]:=[[Π_w∈D^→k_0,w]]^Z_q for each set of a positive integer satisfying η<┌log L┐ and D satisfying D⊆Z_L and 2^η+1≤|D|≤min(2^η+1, L) (“2:” to “5:” under Scheme 5). Here, assuming that x^A is an arbitrary real number, ┌x^A┐ is a ceiling function. That is, ┌x^A┐ represents a minimum integer of x^A or more for the real number x^A. Here, “:=” has the same meaning as “=”.

In “6:” to “8:” under Scheme 5-1, the first apparatus, the second apparatus and the third apparatus 3 perform a process of calculating [[^→f_j]]^Z_q:=Π_j<L[[^→k′_0,w]] for each j where j<2^L (“6:” to “8:” under Scheme 5-1). Here, assuming that a wth bit of bit expression of j is j_w,^→k′_0,w is ^→k′_0,w=1−^→k_0,w when j_w=0 and ^→k′_0,w=^→k_0,w when j_w=1.

In “9:” under Scheme 5-1, the first apparatus, the second apparatus and the third apparatus 3 perform a process of calculating [[(^→f_j)_u]]^Z_q:=Σ_0≤t<u[[(^→f_j)_t]]^Z_q+[[s_j]]^Z_q (“9:” under Scheme 5-1). Here, it is assumed that an arbitrary vector is ^→f^A, (^→f^A)_u is a uth element of the vector ^→f^A, s₀:=0, s_j:=Σ_0≤u<m(^→f_j−1)_u+s_j−1 for j=1, . . . , 3, and the number of elements of vector ^→f_j−1 is m.

In “10:” under Scheme 5-1, the first apparatus, the second apparatus and the third apparatus 3 perform a process of calculating [σ₀⁻¹]^Z_q,01:=Σ_{0≤j<2{circumflex over ( )}L}[^→f′_j^→f_j]^Z_q,01 using (2,2) output product sum (“10:” under Scheme 5-1). The (2,2) output product sum can be calculated under Scheme 5-4 in FIG. 14, for example. The (2,2) output product sum is a product sum process to obtain an output in a (2,2)-additive secret variance format.

In “11:” under Scheme 5-1, a process of calculating <σσ₀⁻¹>²⁰:={π}^01,12[(σ₀⁻¹]^Z_q,01 is performed (“11:” under Scheme 5-1). Here, <πσ_i⁻¹>²⁰ represents that πσ_i⁻¹ is shared by the third apparatus and the first apparatus. The process in “11:” under Scheme 5-1 can be implemented by performing a process under Scheme 5-5 in FIG. 15.

Hereinafter, Scheme 5-2 in FIG. 11 and FIG. 12 will be described.

In “1:” under Scheme 5-2, the first, second and third apparatuses perform a process of converting {^→k_i,0}^Z_2, {^→k_i,1}^Z_2, . . . , {^→k_{i, L−1}}^Z_2 to (2,2) additive secret variance [^→k_i,0]^Z_2,01, [^→k_i,1]^Z_2,01, . . . , [^→k_{i, L−1}]^Z_2,01 (“1:” under Scheme 5-2).

In “2:” under Scheme 5-2, the first, second and third apparatuses perform a process of obtaining [^→b₀]^Z_2,20:=[σ_i−1k_i,0]^Z_2,20, [^→b₁]^Z_2,20:=[σ_i−1k_i,1]^Z_2,20, . . . , [^→b_L−1]^Z_2,20:=[σ_i−1k_{i, L−1}]^Z_2,20 by inversely applying {σ_i−1⁻¹} to [^→k_i,0]^Z_2,01, [^→k_i,1]^Z_2,01 . . . , [^→k_{i, L−1}]^Z_2,01 (“2:” under Scheme 5-2).

In “3:” under Scheme 5-2, the first, second and third apparatuses perform a process of converting, [^→b₀]^Z_2,20, [^→b₁]^Z_2,20, . . . , [^→b_L−1]^Z_2,20 to [[b₀]]^Z_q, [[^→b₁]]^Z_q, . . . , [^→b_L−1]^Z_q through mod 2→mod q conversion (“3:” under Scheme 5-2).

In “4:” to “7:” under Scheme 5-2, the first, second and third apparatuses perform a process of calculating [[K_D]]:=[[Π_w∈D^→k_0,w]]^Z_q on each set of a positive integer η satisfying η<┌Flog L┐ and D satisfying D⊆Z_L and 2^η1≤|D|≤min(2^η+1, L) (“4:” to “7:” under Scheme 5-2).

In “8:” to “10:” under Scheme 5-2, the first, second and third apparatuses perform a process of calculating [[^→f_j]]^Z_q:=Π_j<1, [[^→k′_i,w]] for each j where j<2^L(“8:” to “10:” under Scheme 5-2). Here, it is assumed that the wth bit of bit expression of j is j_w, ^→k′_i,w is ^→k′_i,w=1−^→k_i,w when j_w=0 and ^→k′_i,w=^→k_i,w when j_w=1.

In “11:” under Scheme 5-2, the first, second and third apparatuses perform a process of calculating [[(^→f′_j)_u]]^Z_q:=Σ_0≤t<u[[(^→f_j)_t]]^Z_q+[[s_j]]^Z_q (“11:” under Scheme 5-2).

In “12:” under Scheme 5-2, the first, second and third apparatuses perform a process of calculating [σ′⁻¹]^Z_q,20:=Σ_{0≤j<2{circumflex over ( )}L}[^→f′_j^→f_j]^Z_q,20 through (2,2) output product sum (“11:” under Scheme 5-2).

In “13:” under Scheme 5-2, the first, second and third apparatuses perform a process of obtaining [σ⁻¹]^Z_q,01:=[(σ_i−1⁻¹σ′⁻¹]^Z_q,01 by applying {σ_i−1⁻¹} to [σ⁻¹]^Z_q,20 (“13:” under Scheme 5-2).

In “14:” under Scheme 5-2, the first, second and third apparatuses perform a process of calculating <πσ_i⁻¹>²⁰:={π}^01,12[σ⁻¹]^Z_q,01 (“14:” under Scheme 5-2). Note that the process in “14:” under Scheme 5-2 can be implemented by performing a process under Scheme 5-5 in FIG. 15.

Note that in the case of L=2, Scheme 5-1 and Scheme 5-2 become Scheme 5-1 and Scheme 5-2 described in FIG. 16, FIG. 17 and FIG. 18 respectively.

Using inverse substitution in this way, it is possible to make the algorithm of confidential sort simpler than in the prior art. This allows confidential sort to be performed faster than in the prior art.

[Modifications]

As will be described below, optimization using communication channels may be used. In other words, the communication channels may be effectively used so that there are as few empty communication channels as possible.

For example, the process of mod 2 to mod q in “3:” the process of multiplication in “4:” and the process of (2,2) output product sum in “7:” under Scheme 5-1 and Scheme 5-2 in FIG. 16, FIG. 17 and FIG. 18 have a degree of freedom in communication directions. These processes may be performed by effectively using the communication channels so that there are as few empty communication channels as possible.

When L′ is not a multiple of L, the key having length L′ may be divided so that NL+L″=L′. The key having length L′ may be divided so that Σ_i=0^N−1L_i=L′. In this case, for example, ^→k_n becomes a vector constructed of an (n+1)th (n=0, . . . , N−1) bit string when the key having length L′ is divided so that Σ_i=0^N−1L_i=L′. In this way, N may be assumed to be a predetermined positive integer and the (n+1)th (n=0, . . . , N−1) bit string when the key is divided into N bit strings may be assumed to be ^→k_n.

In addition, it goes without saying that changes can be made as appropriate without departing from the spirit and scope of the present invention.

[Program and Recording Medium]

When, for example, a process in each apparatus is implemented by a computer, process contents of a function that should be possessed by each part of each apparatus are written by a program. By causing a computer to execute this program, processes of the respective apparatus are implemented on the computer.

The program in which the process contents are written can be recorded in a computer-readable recording medium. As the computer-readable recording medium, any magnetic recording apparatus, optical disk, magnetooptical recording medium, semiconductor memory or the like may be used.

Processes of the respective parts may be configured by causing a predetermined program to be executed on a computer or at least some of the processes may be implemented by hardware.

Claims

1. A confidential sort system comprising first, second, . . . , and Mth apparatuses, wherein assuming that a body of data to be sorted is a value →v, a tag that determines order after the sort is a key, a bit length of the key is L′, N is a predetermined positive integer, an (n+1)th (n=0, . . . , N−1) bit string when the key is divided into N bit strings is →kn, an arbitrary value or substitution is xA, {xA} is a replicated secret variance of xA, an arbitrary value is xA, is a secret variance having homomorphism of xA, and M is a predetermined positive integer of 2 or more,the first, second, . . . , and Mth apparatuses obtain inverse substitution [[σ0−1]] of L-bit stable sort of {→k0},the first, second, . . . , and Mth apparatuses perform, on i=1, . . . , N−1:a process of obtaining {σi−1−1} by converting [[σi−1−1]] to hybrid substitution;a process of obtaining {σi−1→ki} by inversely substituting {→ki} using {σi−1−1};a process of obtaining inverse substitution [[σ′i−1]] of L-bit stable sort of [[σi−1→ki]];a process of obtaining [[σi−1]]:=[[σi−1−1σ′1−1]] by synthesizing {σi−1−1} with [[σ′i−1]]; anda process of obtaining {σN−1−1} by converting [[σN−1−1]] to hybrid substitution, andthe first, second, . . . , and Mth apparatuses output [[σN−1→v]] by inversely substituting [[→v]] using {σN−1−1}.

2. A confidential sort system comprising first, second, . . . , and Mth apparatuses, wherein assuming that a body of data to be sorted is a value →v, a tag that determines order after the sort is a key, a bit length of the key is L′, N is a predetermined positive integer, an (n+1)th (n=0, . . . , N−1) bit string when the key is divided into N bit strings is →kn, an arbitrary value or substitution is xA, {xA} is a replicated secret variance of xA, an arbitrary value is xA, [[xA]] is a secret variance having homomorphism of xA, <xA> is a semi-public value, [xA] is a (2,2) additive secret variance, {→ki}(Z_2){circumflex over ( )}L=({→ki,0}Z_2, {→ki,1}Z_2, . . . , {→ki, L−1}Z_2)(i=0, . . . , N−1), q is a predetermined positive integer, (→fA)u is a uth element of vector →fA where →fA is an arbitrary vector, s0:=0, sj:=Σ0≤u<m(→fj−1)u+sj−1 for j=1, . . . , 3, the number of elements of vector →fj−1 is m, 01 represents a set of the first and second apparatuses, 12 represents a set of the second and third apparatuses, 20 represents a set of the third and first apparatuses, G is an arbitrary group, ring or body, P is a set of apparatuses, [xA]G,P represents a share of [xA] represented on G of P, π is predetermined substitution, {π}P means a sub-share of {π} shared by P, {π}01,12 represents a replicated secret variance in which π={π}12{π}01, and <πσi−1>20 shows that πσi−1 is shared by the third and first apparatuses,the first, second and third apparatuses perform:a process of converting, through mod 2→mod q conversion, {→k0,0}Z_2, {→k0,1}Z_2, . . . , {→k0, L−1}Z_2 to [[→k0,0]]Z_q, [[→k0,1]]Z_q, . . . , [[→k0, L−1]]Z_q,a process of calculating [[KD]]:=[[Πw∈D→k0,w]]Z_q for each set of a positive integer satisfying η<┌log L┐ and D satisfying D⊆ZL and 2η+1≤|D|≤min(2η+1, L);a process of calculating [[→fj]]Z_q:=Πj<L[[k′0,w]] for each j where j<2L assuming the wth bit of bit expression of j as jw and →k′0,w is →k′0,w=1−→k0,w when jw=0 and →k′0,w=→k0,w when jw=1;a process of calculating [[(→f′j)u]]Z_q:=Σ0≤t<u[[(→fj)t]]Z_q+[[sj]]Z_q;a process of calculating [σ0−1]Z_q,01:=Σ0≤j<2{circumflex over ( )}L[→f′j→fj]Z_q,01 through (2,2) output product sum operation; anda process of calculating <πσ0−1>20:={π}01,12[σ0−1]Z_q,01,the first, second and third apparatuses perform:a process of converting {→ki,0}Z_2, {→ki,1}Z_2, . . . , {→ki, L−1}Z_2 to (2,2) additive secret variance [→ki,0]Z_2,01, [→ki,1]Z_2,01, . . . , [→ki, L−1]Z_2,01 assuming i=1, . . . , N−1;a process of inversely applying {σi−1−1} to [→ki,0]Z_2,01, [→ki,1]Z_2,01 . . . , [→ki, L−1]Z_2,01 to obtain [→b0]Z_2,20:=[σi−1ki,0]Z_2,20, [→b1]Z_2,20:=[σi−1ki,1]Z_2,20, . . . , [→bL−1]Z_2,20:=[σi−1ki, L−1]Z_2,20;a process of converting [→b0]Z_2,20, [→b1]Z_2,20, . . . , [→bL−1]Z_2,20 to [[→b0]]Z_q, [[→b1]]Z_q, . . . , [→bL−1]Z_q through mod 2→mod q conversion;a process of calculating [[KD]]:=[[Πw∈D→k0,w]]Z_q for each set of a positive integer satisfying η<┌log L┐ and D satisfying D⊆ZL and 2η+1≤|D|≤min(2η+1, L);a process of calculating [[→fj]]Z_q:=Πj<L[[→k′i,w]] for each j where j<2L assuming the wth bit of bit expression of j is jw, →k′i,w is →k′i,w=1−→ki,w when jw=0 and →k′i,w=→ki,w when jw=1;a process of calculating [[(→f′j)u]]Z_q:=Σ0≤t<u[[(→fj)t]]Z_q+[[sj]]Z_q;a process of calculating [σ′−1]Z_q,20:=Σ0≤j<2{circumflex over ( )}L[→f′j→fj]Z_q,20 through (2,2) output product-sum operation;a process of applying to {σi−1−1} to obtain [σ−1]Z_q,20 to obtain [σ−1]Z_q,01:=[σi−1−1σ′−1]Z_q,01; anda process of calculating <πσi−1>20:={π}01,12[σ−1]Z_q,01 on i=1, . . . , N−1, andthe first, second, . . . , and Mth apparatuses inversely substitute [[→v]] with {σ−1} assuming that {σ−1}=(<πσN−1−1>20, {π}01,12) and output [[σ→v]].

3. A confidential sort method based on assumption that a body of data to be sorted is a value →v, a tag that determines order after the sort is a key, a bit length of the key is L′, N is a predetermined positive integer, an (n+1)th (n=0, . . . , N−1) bit string when the key is divided into N bit strings is →kn, an arbitrary value or substitution is xA, {xA} is a replicated secret variance of xA, an arbitrary value is xA, [[xA]] is a secret variance having homomorphism of xA, and M is a predetermined positive integer of 2 or more, the method comprising: a step in which first, second, . . . , and Mth apparatuses obtain inverse substitution [[σ0−1]] of L-bit stable sort of {→k0},a step in which the first, second, . . . , and Mth apparatuses perform, on i=1, . . . , N−1:a process of obtaining {σi−1−1} by converting [[σi−1−1]] to hybrid substitution;a process of obtaining {σi−1→ki} by inversely substituting using {σi−1−1};a process of obtaining inverse substitution [[σ′i−1]] of L-bit stable sort of [[σi−1→ki]];a process of obtaining [[σ′i−1]]:=[[σi−1−1σ′i−1]] by synthesizing {σ′i−1} with [[σ′i−1]]; anda process of obtaining {σN−1−1} by converting [[σN−1−1]] to hybrid substitution, anda step in which the first, second, . . . , and Mth apparatuses output [[σN−1→v]] by inversely substituting [[43 v]] using {σN−1−1}.

4. A confidential sort method based on assumption that a body of data to be sorted is a value →v, a tag that determines order after the sort is a key, a bit length of the key is L′, N is a predetermined positive integer, an (n+1)th (n=0, . . . , N−1) bit string when the key is divided into N bit strings is →kn, an arbitrary value or substitution is xA, {xA} is a replicated secret variance of xA, an arbitrary value is xA, [[xA]] is a secret variance having homomorphism of xA, <xA> is a semi-public value, [xA] is a (2,2) additive secret variance, {→ki}(Z_2){circumflex over ( )}L=({→ki,0}Z_2, {→ki,1}Z_2, . . . , {→ki, L−1}Z_2) (i=0, . . . , N−1), q is a predetermined positive integer, (→fA)u is a uth element of vector →fA where →fA is an arbitrary vector, s0:=0, sj:=Σ0≤u<m(→fj−1)u+sj−1 for j=1, . . . , 3, the number of elements of vector →jj−1 is m, 01 represents a set of the first and second apparatuses, 12 represents a set of the second and third apparatuses, 20 represents a set of the third and first apparatuses, G is an arbitrary group, ring or body, P is a set of apparatuses, [xA]G,P represents a share of [xA] represented on G of P, π is predetermined substitution, {π}P means a sub-share of {π} shared by P, {π}01,12 represents a replicated secret variance in which π={π}12{π}01, and <πσi−1>20 shows that πσi−1 is shared by the third and first apparatuses, the method comprising: a step in which the first, second and third apparatuses perform:a process of converting, through mod 2→mod q conversion, {→k0,0}Z_2, {→k0,1}Z_2, . . . , {→k0, L−1}Z_2 to [[→k0,0]]Z_q, [[→k0,1]]Z_q, [[→k0,1]]Z_q, . . . , [[→k0, L−1]]Z_q;a process of calculating [[KD]]:=[[Πw∈D→k0,w]]Z_q for each set of a positive integer η satisfying η<┌log L┐ and D satisfying D⊆ZL and 2η+1≤|D|≤min(2η+1, L);a process of calculating [[→fj]]Z_q:=Πj<L[[→k′0,w]] for each j where j<2L assuming the wth bit of bit expression of j as jw and →k′0,w is →k′0,w=1−→k0,w when jw=0 and →k′0,w=→k0,w when jw=1;a process of calculating [[→fj)u]]Z_q:=Σ0≤t<u[[→fj)t]]Z_q+[[sj]]Z_q;a process of calculating [σ0−1]Z_q,01:=Σ0≤j<2{circumflex over ( )}L[→f′j→fj]Z_q,01 through (2,2) output product sum operation; anda process of calculating <πσ0−1>20:={π}01,12[π0−1]Z_q,01,the first, second and third apparatuses perform:a process of converting {→ki,0}Z_2, {→ki,1}Z_2, . . . , {→ki, L−1}Z_2 to (2,2) additive secret variance [→ki,0]Z_2,01, [→ki,1]Z_2,01, . . . , [→ki, L−1]Z_2,01 assuming i=1, . . . , N−1;a process of inversely applying {σi−1−1} to [→ki,0]Z_2,01, [→ki,1]Z_2,01 . . . , [→ki, L−1]Z_2,01 to obtain [→b0]Z_2,20:=[σi−1ki,0]Z_2,20, [→b1]Z_2,20:=[σi−1ki,1]Z_2,20, . . . , [→bL−1]Z_2,20:=[σi−1ki, L−1]Z_2,20;a process of converting [→b0]Z_2,20, [→b1]Z_2,20, . . . , [→bL−1]Z_2,20 to [[→b0]]Z_q, [[→b1]]Z_q, . . . , [→bL−1]Z_q through mod 2→mod q conversion;a process of calculating [[KD]]:=[[Πw∈D→k0,w]]Z_q for each set of a positive integer η satisfying η<┌log L┐ and D satisfying D⊆ZL and 2η+1≤|D|≤min(2η+1, L);a process of calculating [[→fj]]Z_q:=Πj<L[[→k′i,w]] for each j where j<2L assuming the wth bit of bit expression of j is jw, →k′i,w is →k′i,w=1−→ki,w when jw=0 and →k′i,w=→ki,w when jw=1;a process of calculating [[(→f′j)u]]Z_q:=Σ0≤t<u[[(→fj)t]]Z_q+[[sj]]Z_q;a process of calculating [σ′−1]Z_q,20:=Σ0≤j<2{circumflex over ( )}L[→f′j→fj]Z_q,20 through (2,2) output product-sum operation;a process of applying {σi−1−1} to [σ−1]Z_q,20 to obtain [σ−1]Z_q,01:=[σi−1−1σ′−1]Z_q,01; anda process of calculating <πσi−1>20:={π}01,12[σ−1]Z_q,01 on i=1, . . . , N−1, anda step in which the first, second, . . . , and Mth apparatuses inversely substitute [[→v]] with {σ−1} assuming that {σ−1}=(<πσN−1−1>20, {π}01,12) and output [[σ→v]].