Verifiable evaluations of machine learning models using zkSNARKs

TL;DR

This paper introduces a method using zkSNARKs to create verifiable proofs of machine learning model evaluations, enabling users to confirm model performance and fairness without re-running benchmarks.

Contribution

It presents a novel approach for verifiable model evaluation using zero-knowledge proofs, applicable to standard neural networks with varying computational costs.

Findings

Successfully demonstrated verifiable evaluations on real-world models

Identified key challenges and solutions in implementing zkSNARKs for ML

Provides a new transparency paradigm for private model assessments

Abstract

In a world of increasing closed-source commercial machine learning models, model evaluations from developers must be taken at face value. These benchmark results-whether over task accuracy, bias evaluations, or safety checks-are traditionally impossible to verify by a model end-user without the costly or impossible process of re-performing the benchmark on black-box model outputs. This work presents a method of verifiable model evaluation using model inference through zkSNARKs. The resulting zero-knowledge computational proofs of model outputs over datasets can be packaged into verifiable evaluation attestations showing that models with fixed private weights achieve stated performance or fairness metrics over public inputs. We present a flexible proving system that enables verifiable attestations to be performed on any standard neural network model with varying compute requirements. For the first time, we demonstrate this across a sample of real-world models and highlight key challenges and design solutions. This presents a new transparency paradigm in the verifiable evaluation of private models.