Recently, motivated by its increased use in real-world applications, there has been a growing interest on the reduction of trust in the generation of the common reference string (CRS) for zero-knowledge (ZK) proofs. This line of research was initiated by the introduction of subversion non-interactive ZK (NIZK) proofs by Bellare et al. (ASIACRYPT’16).
Here, the zero-knowledge property needs to hold even in case of a malicious generation of the CRS. Groth et al. (CRYPTO’18) then introduced the notion of updatable zk-SNARKS, later adopted by Lipmaa (SCN’20) to updatable quasi-adaptive NIZK (QA-NIZK) proofs. In contrast to the subversion setting, in the updatable setting one can achieve stronger soundness guarantees at the cost of reintroducing some trust, resulting in a model in between the fully trusted CRS generation and the subversion setting.
It is a promising concept, but all previous updatable constructions are ad-hoc and tailored to particular instances of proof systems. Consequently, it is an interesting question whether it is possible to construct updatable ZK primitives in a more modular way from simpler building blocks.