Certified Everlasting Zero-Knowledge Proof for QMA

Abstract

In known constructions of classical zero-knowledge protocols for ${\textbf {NP}}$, either zero-knowledge or soundness holds only against computationally bounded adversaries. Indeed, achieving both statistical zero-knowledge and statistical soundness at the same time with classical verifier is impossible for ${\textbf {NP}}$ unless the polynomial-time hierarchy collapses, and it is also believed to be impossible even with a quantum verifier. In this work, we introduce a novel compromise, which we call the certified everlasting zero-knowledge proof for ${\textbf {QMA}}$. It is a computational zero-knowledge proof for ${\textbf {QMA}}$, but the verifier issues a classical certificate that shows that the verifier has deleted its quantum information. If the certificate is valid, even an unbounded malicious verifier can no longer learn anything beyond the validity of the statement.

We construct a certified everlasting zero-knowledge proof for ${\textbf {QMA}}$. For the construction, we introduce a new quantum cryptographic primitive, which we call commitment with statistical binding and certified everlasting hiding, where the hiding property becomes statistical once the receiver has issued a valid certificate that shows that the receiver has deleted the committed information. We construct commitment with statistical binding and certified everlasting hiding from quantum encryption with certified deletion by Broadbent and Islam [TCC 2020] (in a black-box way), and then combine it with the quantum sigma-protocol for ${\textbf {QMA}}$ by Broadbent and Grilo [FOCS 2020] to construct the certified everlasting zero-knowledge proof for ${\textbf {QMA}}$. Our constructions are secure in the quantum random oracle model. Commitment with statistical binding and certified everlasting hiding itself is of independent interest, and there will be many other useful applications beyond zero-knowledge.