Score:1

ddh and statistical distance

hm flag

Let $\mathbb{G}$ be a cyclic group of prime order q and generated by g. Let $D$ be the uniform distribution over $\mathbb{G}^3$. Let $D_{dh}$ be the uniform distribution over the set of all DH-triples $(g^{\alpha}, g^{\beta}, g^{\alpha\beta})$. Let $D_{ndh}$ be the uniform distribution over the set of all non-DH-triples $(g^{\alpha}, g^{\beta}, g^{\gamma})$ with $\gamma\neq\alpha\beta$. Answer the following questions:

(a) Show that the statistical distance between $D$ and $D_{ndh}$ is $1/q$

(b) Show that, under DDH assumption, the distributions $D_{dh}$ and $D_{ndh}$ are computationally indistinguishable.

What I did till now is trying to show (a) $SD(D,D_{ndh})=\frac{1}{2}\sum_{a,b,c}|(Pr[D=(a,b,c)]-Pr[D_{ndh}=(a,b,c)]|=\frac{1}{2}q^3|\frac{1}{q^3}-\frac{1}{q^2}\frac{1}{q-1}|=\frac{1}{q-1}$

I got to this point thinking that every truple is equally possible over $\mathbb{G^3}$ and that g is a generator so there must exist a truple $(x,y,z)$ such that $(g^x,g^y,g^z)=(a,b,c)$ just considering that $z$ must be different from $xy$.

For point (b) I actually don't know how to procede, I know I should show that no PPT adversary could distinguish between the two, and it is trivially DDH and so I don't what I should prove, maybe doing a reduction to DDH?

Marc Ilunga avatar
tr flag
Hint for b) Look into the relationship between statistical distance and distinguishing advantage. for a) The question has $1/q$ whereas you find $1/(q-1)$.
Cristie avatar
hm flag
For a) I know what I did is not the answer, that's why I asked
Daniel S avatar
ru flag
HINT: For a) it reads that you think that the probability that sampling $D_{ndh}$ returns the triple $(a,b,c)$ is $1/q^2(q-1)$ for all triples. In fact it is 0 for some triples and non-zero for others.
Daniel S avatar
ru flag
HINT: For part b) work by contradiction: suppose that you did have a computational means of distinguishing $D_{dh}$ and $D_{ndh}$, how could you use this to construct a DDH adversary?
Marc Ilunga avatar
tr flag
Upon re-reading the question,I think I misread it I initially including what you were doing for a). So ignore my comments.
I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.