Abstract:The widespread adoption of machine learning (ML) in various critical applications, from healthcare to autonomous systems, has raised significant concerns about privacy, accountability, and trustworthiness. To address these concerns, recent research has focused on developing zero-knowledge machine learning (zkML) techniques that enable the verification of various aspects of ML models without revealing sensitive information. Recent advances in zkML have substantially improved efficiency; however, these efforts have primarily optimized the process of proving ML computations correct, often overlooking the substantial overhead associated with verifying the necessary commitments to the model and data. To address this gap, this paper introduces two new Commit-and-Prove SNARK (CP-SNARK) constructions (Apollo and Artemis) that effectively address the emerging challenge of commitment verification in zkML pipelines. Apollo operates on KZG commitments and requires white-box use of the underlying proof system, whereas Artemis is compatible with any homomorphic polynomial commitment and only makes black-box use of the proof system. As a result, Artemis is compatible with state-of-the-art proof systems without trusted setup. We present the first implementation of these CP-SNARKs, evaluate their performance on a diverse set of ML models, and show substantial improvements over existing methods, achieving significant reductions in prover costs and maintaining efficiency even for large-scale models. For example, for the VGG model, we reduce the overhead associated with commitment checks from 11.5x to 1.2x. Our results suggest that these contributions can move zkML towards practical deployment, particularly in scenarios involving large and complex ML models.

Multivariate Multi-Polynomial Commitment and its Applications

A Survey of Multivariate Polynomial Commitment Schemes

Zero-Knowledge MIPs using Homomorphic Commitment Schemes

Zeromorph: Zero-Knowledge Multilinear-Evaluation Proofs from Homomorphic Univariate Commitments

Artemis: Efficient Commit-and-Prove SNARKs for zkML

Efficient Non-Interactive Zero Knowledge Arguments for Set Operations.

Cuzk: Accelerating Zero-Knowledge Proof with A Faster Parallel Multi-Scalar Multiplication Algorithm on GPUs.

On the Security of KZG Commitment for VSS

Info-Commit: Information-Theoretic Polynomial Commitment

Transparent Polynomial Delegation and Its Applications to Zero Knowledge Proof

Two prover perfect zero knowledge for MIP*

Improvements on Non-Interactive Zero-Knowledge Proof Systems Related to Quadratic Residuosity Languages.

Compact Proofs of Partial Knowledge for Overlapping CNF Formulae

MP-SPDZ: A Versatile Framework for Multi-Party Computation

Secure Multiparty Computation of Set Membership and Its Applications

Collaborative CP-NIZKs: Modular, Composable Proofs for Distributed Secrets

Simple Matrix - A Multivariate Public Key Cryptosystem (mpkc) for Encryption

Multivariate Oblivious Polynomial Evaluation Protocol

GZKP: A GPU Accelerated Zero-Knowledge Proof System

Publicly Auditable MPC-as-a-Service with succinct verification and universal setup