Prover-efficient commit-and-prove zero-knowledge SNARKs

Zk-SNARKs (succinct non-interactive zero-knowledge arguments of knowledge) are needed in many applications. Unfortunately, all previous zk-SNARKs for interesting languages are either inefficient for prover, or are non-adaptive and based on an commitment scheme that does depend both on the prover’s input and on the language, i.e., they are not commit-and-prove (CaP) SNARKs. We propose a prooffriendly extractable commitment scheme, and use it to construct proverefficient adaptive CaP succinct zk-SNARKs for different languages, that can all reuse committed data. In new zk-SNARKs, the prover computation is dominated by a linear number of cryptographic operations. We use batch-verification to decrease the verifier’s computation.