Min-round resettable zero-knowledge in the public-key model

Abstract

In STOC 2000, Canetti, Goldreich, Goldwasser, and Micali put forward the strongest notion of zero-knowledge to date, resettable zero-knowledge (RZK) and implemented it in constant rounds in a new model, where the verifier simply has a public key registered before any interaction with the prover. To achieve ultimate round efficiency, we advocate a slightly stronger model. Informally, we show that, as long as the honest verifier does not use a given public key more than a fixed-polynomial number of times, there exist 3-round (which we prove optimal) RZK protocols for all of NP.