22.04.2014 Views

a590003

a590003

a590003

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

f<br />

$<br />

←<br />

P t (p)<br />

r ′ a = (f(·, a), f(a, ·))<br />

r a [a ′ ] = s a ′[a]<br />

r a [h] = f(a, h)<br />

G[h, h ′ ] = (p > ⊥)<br />

G[h, a] = eq(s a [h], f(h, a))<br />

G[a, j] = G a [j]<br />

p h = p(h) .<br />

Notice that this is exactly how p h is defined by the VSS functionality. So, in order to prove security,<br />

it is enough to give a simulator Sim that on input p A , s A , G A , outputs G, r A and r<br />

A ′ as defined in<br />

the above system of equations. See Figure 20 (right).<br />

The problem faced by the simulator is that it cannot test p > ⊥ and generate f as in the equations<br />

because it does not know the value of p, rather it only has partial information p A = p(A). The<br />

first condition p > ⊥ is easy to check because it is equivalent to p a = p(a) > ⊥ for any a ∈ A. In<br />

order to complete the simulation, we observe that the equations only depend on the 2t polynomials<br />

f(·, A) and f(A, ·). The next lemma shows that, given p(A), the polynomials f(·, A) and f(A, ·) are<br />

statistically independent from p, and their distribution can be easily sampled.<br />

$<br />

Lemma 2 Let p ∈ F t [X], let f ← P t (p), and for all a ∈ A, let g a = f(·, a) and h a = f(a, ·). The<br />

conditional distribution of (g a , h a ) a∈A given p(A) is statistically independent of p, and it can be<br />

generated by the following algorithm Samp(p A ): first pick random polynomials h a ∈ F t [Y ] independently<br />

and uniformly at random subject to the constraint h a (0) = p a . Then, pick g a ∈ F t [X], also<br />

independently and uniformly at random, subject to the constraint g a (A) = h A (a).<br />

Using the algorithm from the lemma, we obtain the following simulator Sim:<br />

Sim(p A , s A , G A ) = (G A , r A , r ′ A ):<br />

(g A , h A ) ← Samp(p A )<br />

r ′ A = (g A, h A )<br />

r a [h] = h a (h) (h ∈ H, a ∈ A)<br />

r a [a ′ ] = s a ′[a] (a, a ′ ∈ A)<br />

G[h, h ′ ] = ∨ a∈A (p a > ⊥) (h, h ′ ∈ H)<br />

G[h, a] = eq(s a [h], g a (h)) (h ∈ H, a ∈ A)<br />

G[a, j] = G a [j]<br />

(a ∈ A, j ∈ [n])<br />

As usual, if p = ⊥, then p A = ⊥ A and by convention Samp(p A ) = {⊥, ⊥}.<br />

Dishonest dealer security. We now look at the case where the dealer is not honest. As above,<br />

for all A ⊆ [n] where |A| = t and n ≥ 4t + 1, define H = [n] \ A. When the players in the set<br />

A are corrupted (and thus the players in H are honest), an execution of the VSS protocol with<br />

dishonest dealer is given by the system (Player[H] | Net’ | Net | Graph) with inputs s ′ , r A , G A ,<br />

and outputs r<br />

A ′ , s A, p H and G A . As above, we start with an equational description of the system,<br />

and will simplify it below into a form where the construction of a corresponding simulator becomes<br />

obvious. For all i, j ∈ [n], h, h ′ ∈ H, and a ∈ A, we have<br />

(g h , h h ) = s ′ [h]<br />

r a ′ = s ′ [a]<br />

r i [h] = g h (i)<br />

r i [a] = s a [i]<br />

G[h, h ′ ] = eq(g h ′(h), h h (h ′ ))<br />

G[h, a] = eq(s a [h], h h (a))<br />

G[a, j] = G a [j]<br />

∨ [<br />

o h =<br />

cliqueC (G) ∧ interpolate C,t (r h ) ]<br />

C⊆[n],|C|≥n−t<br />

p h = o h (0) .<br />

25<br />

12. An Equational Approach to Secure Multi-party Computation

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!