Ensuring Privacy and Integrity in Dynamic Settings: Updatable Vector Commitments and Zero-Knowledge Sets

In this work we consider the common scenario where a server first commits to a vector or set of confidential elements and then efficiently answers queries on the committed elements. Interestingly, we show that the case in which the elements change over time has been overlooked by the scientific literature that investigated Vector Commitments and Zero-Knowledge Sets. Next, we provide natural definitions addressing confidentiality with updates and we show limitations of well-known constructions. We demonstrate that a folklore transform of Catalano and Fiore [PKC 2013] to add hiding to a vector commitment fails according to our definition. On the positive side, on top of a recent construction of an updatable vector commitment called Reckle Tree [CCS 2024] that does not satisfy confidentiality, we provide constructions of Updatable Vector Commitments and Zero-Knowledge Sets along with an experimental validation.