Normal Alignment: Improved Cryptanalytic Sign Recovery on Hard-Label Networks
Researchers propose Normal Alignment, improving cryptanalytic sign recovery for hard-label neural networks and enabling polynomial-time full model extraction.
The paper improves on Carlini et al.'s EUROCRYPT 2025 cryptanalytic extraction of hard-label (S1) DNNs, whose Future Toggle sign-recovery method offered only marginal advantage over random guessing and triggered exponential-time enumeration on errors. Normal Alignment infers neuron signs via expected length differences between projected normals of adjacent decision facets at dual points, delivering higher voting accuracy and low-confidence errors. Combined with the SOE extension, it achieves exact polynomial-time full sign recovery: CIFAR-10 (192-64x8-10) and MNIST (64-96x3-32-10) models are fully recovered where the prior method required 2^52 or 2^82 sign guesses.
- Normal Alignment outperforms the Future Toggle sign-recovery method on hard-label DNNs
- Enables polynomial-time full sign recovery on CIFAR-10 and MNIST models
- Prior approach required exponential enumeration of 2^52 to 2^82 guesses
- Builds on Carlini et al.'s EUROCRYPT 2025 model extraction breakthrough
- Products
- Normal Alignment
From the original article
1 min read at arxiv.orgAt EUROCRYPT 2025, Carlini et al. proposed a breakthrough in the cryptanalytic extraction on hard-label (S1) deep neural networks (DNNs), demonstrating polynomial-time signature and sign recovery. However, Carlini et al.'
By Shi Tang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.