Hard-Label Cryptanalytic Extraction of Neural Network Models

TL;DR

This paper introduces the first theoretical attack capable of extracting neural network parameters under the hard-label setting, validated on real datasets with practical efficiency.

Contribution

It presents a novel attack method that achieves functionally equivalent extraction in the hard-label setting for ReLU neural networks, a previously unresolved challenge.

Findings

Effective extraction on networks with 10^5 parameters within hours

Validated on MNIST and CIFAR10 datasets

Works under realistic hard-label access constraints

Abstract

The machine learning problem of extracting neural network parameters has been proposed for nearly three decades. Functionally equivalent extraction is a crucial goal for research on this problem. When the adversary has access to the raw output of neural networks, various attacks, including those presented at CRYPTO 2020 and EUROCRYPT 2024, have successfully achieved this goal. However, this goal is not achieved when neural networks operate under a hard-label setting where the raw output is inaccessible. In this paper, we propose the first attack that theoretically achieves functionally equivalent extraction under the hard-label setting, which applies to ReLU neural networks. The effectiveness of our attack is validated through practical experiments on a wide range of ReLU neural networks, including neural networks trained on two real benchmarking datasets (MNIST, CIFAR10) widely used in computer vision. For a neural network consisting of $10^5$ parameters, our attack only requires several hours on a single core.