Open krnak opened 1 year ago
You are correct @krnak. This should be at least partially fixed by PR #148 by @CarloModicaPortfolio which I looked over again this morning.
I am not sure #148 solves the entire blinding issue. I am concerned that more blinding for the z_poly
may be needed. Will you look it over and give a review?
I have reviewed #148. New blinding looks sufficient to me.
Method
StandardComposer::add_blinding_factors
uses 4 random scalars to blind wire polynomialsw_l
andw_r
and the permutation polynomialz_poly
polynomial. Also notice that blinding factors ofw_o
andw_4
does not contribute to a blinding ofz_poly
, because these values will cancel out from equation ofz_poly
.Since the verifier gets 5 evaluations of these 3 polynomials, namely
four scalars are definitely not enough to blind these polynomials. It follows that the current implementation of the protocol does not have the zero-knowledge property.