• No results found

2.5 Complexity Assumptions

In this section, we review the complexity assumptions used throughout this thesis.

2.5.1 Discrete Logarithm Assumption

The discrete logarithm (DL) assumption [Odl85] in a finite field is one of the basic assumptions in cryptography research. The DL assumption is defined as follows.

Definition 2.8 Discrete Logarithm (DL) Assumption [Odl85]. Let G(1`) → (p,G) and G = hgi. Given (g, y) ∈ G2, we say that the discrete logarithm assumption holds on G if no PPT adversary A can compute a x ∈ Zp such that y = gx with the advantage

AdvADL = Pr [y = gx|A(p, g, y,G) → x] ≥ (`)

where the probability is taken over the random choice of y ∈ G and the bits consumed by the adversary A.

2.5.2 Computational Diffie-Hellman Assumption

Diffie and Hellman [DH76] proposed this assumption and constructed a key exchange scheme based on it. This assumption is defined as follows.

Definition 2.9 Computational Diffie-Hellman (CDH) Assumption [DH76]. Let x, y←R Zp, G(1`) → (p,G) and G = hgi. Given (g, gx, gy), we say that the computational Diffie-Hellman assumption holds on G if no PPT adversary A can compute gxy with the advantage

AdvACDH = Pr [A(g, gx, gy) → gxy] ≥ (`)

where the probability is taken over the random choices of x, y ←R Zp and the bits consumed by the adversary A.

Maurer [Mau94] discussed the relationships between DL assumption and CDH assumption.

2.5. Complexity Assumptions 14

2.5.3 Decisional Diflie-Hellman Assumption

Boneh [Bon98] surveyed the various applications of decisional Diffie-Hellman as-sumption and demonstrated some results regarding it security.

Definition 2.10 Decisional Diflie-Hellman (DDH) Assumption [Bon98]. Let x, y, z ←R Zp, G(1`) → (p,G) and G = hgi. Given (g, gx, gy), we say that the decisional Diffie-Hellman assumption holds on G if no PPT adversary A can distinguish (X, Y, Z) = (gx, gy, gxy) from (X, Y, Z) = (gx, gy, gz) with the advantage

AdvADDH = |Pr[A(X, Y, gxy) = 1] − Pr[A(X, Y, gz) = 1]| ≥ (`)

where the probability is taken over the random choices x, y, z ←R Zp and the bits consumed by the adversary A.

2.5.4 Computational Bilinear Diffie-Hellman

Boneh and Franklin [BF01] introduced this assumption. This assumption is as follows.

Definition 2.11 Computational Bilinear Diffie-Hellman (CBDH) Assumption [BF01].

Let GG(1`) → (e, p,G, Gτ) and G = hgi. We say that the computational bilinear Diffie-Hellman assumption holds on (e, p,G, Gτ) if no PPT adversaries A can com-pute e(g, g)abc from (A, B, C) = (ga, gb, gc) with the advantage

AdvACBDH = Pr

A(A, B, C) → e(g, g)abc

≥ (`)

where the probability is taken over the random choices of a, b, c ←R Zp and the bits consumed by A.

2.5.5 Decisional Bilinear Diffie-Hellman Assumption

Boneh and Franklin [BF01] introduced this assumption and used it to construct an identity-based encryption (IBE) scheme. This assumption is defined as follows.

Definition 2.12 Decisional Bilinear Diffie-Hellman (DBDH) Assumption [BF01]. Let a, b, c, z ←R Zp, GG(1`) → (e, p,G, Gτ) andG = hgi. We say that the decisional bilin-ear Diffie-Hellman assumption holds on (p, e,G, Gτ) if no PPT adversary A can dis-tinguish (A, B, C, Z) = (ga, gb, gc, e(g, g)abc) from (A, B, C, Z) = (ga, gb, gc, e(g, g)z) with the advantage

AdvADBDH =

Pr[A(A, B, C, e(g, g)abc) = 1] − Pr[A(A, B, C, e(g, g)z) = 1]

≥ (`)

2.5. Complexity Assumptions 15

where the probability is taken over the random choices of a, b, c, z ←R Zp and the bits consumed by the adversary A.

2.5.6 q-Strong Diffie-Hellman Assumption

Boneh and Boyen [BB04b] proposed this assumption and used it to develop a short signature scheme. This assumption is defined as follows.

Definition 2.13 q-Strong Diffie-Hellman (q-SDH) Assumption [BB04b]. Let x←R Zp, GG(1`) → (p, e, G1,G2,Gτ), G1 = hgi and G2 = hhi. Given a (q + 2)-tuple (g, h, hx, · · · , hxq), we say that the q-strong Diffie-Hellman assumption holds on (p, e,G1,G2,Gτ) if no PPT adversary A can compute (c, gx+c1 ) with the advantage

AdvAq−SDH = Prh

A(g, h, hx, · · · , gxq) → (c, gx+c1 )i

≥ (`)

where c ∈Zp and the probability is taken over the random choice of x←R Zp and the bits consumed by the adversary A.

2.5.7 Chosen-Target Computational Diffie-Hellman Assump-tion

Boldyreva [Bol03] introduced this assumption and used it to design a blind signature scheme. This assumption is defined as follows.

Definition 2.14 Chosen-Target Computational Diffie-Hellman (CT-CDH) Assumption [Bol03]. Let x ←R Zp, G(1`) → (p,G) and G = hgi. Let H : {0, 1} → G be a cryptographic hash function. There are two oracles: target oracle TG(·) and help oracle HG(·). TG(·) takes as input i ∈ Zp and outputs gi ∈G. HG(·) takes as input gi ∈ G and outputs gix ∈ G. Let qT and qH denote the number of times that the two oracles are queried, respectively. We say that the chosen-target computational Diffie-Hellman assumption holds on (p,G) if no PPT adversary A can have the advantage

AdvACT −CDH = Pr

ATG(·),HG(·)(p, H, g, gx) → {(i1, θ1), · · · , (iq+1, θq+1)}

≥ (`) where gij ∈ {g1, g2, · · · , gq+1}, θj = gxij for j = 1, 2, · · · , q + 1, and qH ≤ q < qT.

The CT-CDH assumption is the analogous version of the chosen-target RSA inversion (RSA-CTI) assumption [BNPS02].

2.5. Complexity Assumptions 16

2.5.8 E Xtended Chosen-Target Computational Diffie-Hellman Assumption

Intuitively, in the above CT-CDH assumption, after the adversary A queries the help oracle HG(·) on the elements in G at most qH times, he cannot compute a new element in G to the power of x if its discrete logarithm on the generator and the qH queried elements are unknown. Based on the CT-CDH assumption, we propose the extended CT-CDH (XCT-CDH) assumption, by replacing the target oracle in CT-CDH assumption with (qH + 1) random elements ofG.

Definition 2.15 EXtended Chosen-Target Computational Diffie-Hellman (XCT-CDH) Assumption. Let x←R Zp, G(1`) → (p,G) and G = hgi. Suppose that HG(·) be a help oracle which takes as input gi ∈ G and outputs gix ∈ G. Let qH be the number of the times which the oracle is queried. Given a (q + 1)-tuple (ga1, ga2, · · · , gaq+1), we say that the extended chosen-target computational Diffie-Hellman assumption holds on (p,G) if no PPT adversary A can have the advantage

AdvAXCT −CDH = Pr

AHG(·)(p, g, gx, ga1, · · · , gaq+1) → (gxa1, · · · , gxaq+1)

≥ (`) where ai

R Zp for i = 1, 2, · · · , q + 1 and qH ≤ q.

We have the following result about CT-CDH assumption and XCT-CDH as-sumption.

Theorem 2.1 The extended chosen-target computational Diffie-Hellman assump-tion and the chosen-target Diffie-Hellman assumpassump-tion are equivalent.

Proof: Given the (q+1)-tuple {ga1, ga2, · · · , gaq+1} where aiR Zpfor i = 1, 2, · · · , q+

1, we define a function H : j → gaij ∈ G, where aij ∈ {a1, a2, · · · , aq+1} for j = 1, 2, · · · , q + 1; otherwise H : j → gbj, where bj R

← Zp. So, H(·) is a cryp-tographic hash function 2.6.2, where the domain is Z+ and the range isG.

On the one hand, if an adversary A can break the CT-CDH assumption, we can construct an algorithm B that can use A to break the XCT-CDH assump-tion as follows. Given {ga1, ga2, · · · , gaq+1}, for qT (qT ≤ q + 1) target oracle queries, B responds with gai1, gai2, · · · , gaiqT, where aij ∈ {a1, a2, · · · , aq+1} for j = 1, 2, · · · , qT. For qH (qH ≤ q) help oracle queries, B queries the help or-acle HG(·) in the XCT-CDH assumption, and responds A with {gxai1, gxai2, · · · , gxaiqH}, where aij ∈ {a1, a2, · · · , aq+1} for j = 1, 2, · · · , qH. If A can output