Core Concepts
This work analyzes existing solutions that combine verifiability with privacy-preserving computations over distributed data, in order to preserve confidentiality and guarantee correctness at the same time.
Abstract
The paper presents a comprehensive analysis of existing solutions for Verifiable Privacy-Preserving Computations (VPPCs), which aim to provide both input privacy and public verifiability of the computation results. The authors classify the VPPC schemes into four main classes based on the underlying privacy-preserving computation technique: MPC-based, HE-based, DLT-based, and DP-based.
For each class, the authors discuss the different approaches used to achieve verifiability, such as non-succinct zero-knowledge proofs (ZKPs), succinct ZKPs, homomorphic MACs, and trusted execution environments. They analyze the security, privacy, and public verifiability properties of the schemes, as well as their efficiency and practical aspects.
The key insights from the analysis include:
MPC-based schemes can provide security and public verifiability even when all parties are corrupted, but have higher communication and verification costs.
HE-based schemes can be more efficient for outsourced computations, but require a trusted setup for the zk-SNARK proofs.
DLT-based schemes are suitable for computations with varying participant groups, but are limited by the message size and verification time constraints of the shared ledger.
DP-based schemes are significantly more efficient than the other approaches, but provide weaker privacy guarantees and only approximate correctness.
The authors also identify several underexposed topics, such as the need for input data authentication, reusability of intermediate results, and post-quantum security, which are important for the practical adoption of VPPC schemes.