Paper 2026/2942
Differential–Linear Distinguisher Search via Walsh Spectra over Low-Dimensional Mask Subspaces
Abstract
Searching for differential--linear (DL) distinguishers in ARX and AND--RX ciphers often requires empirical evaluation of multi-round middle correlations. Hourglass(-like) structures provide effective guidance by restricting middle differences and output masks to sparse forms. However, the corresponding low-weight output-mask position need not be the most favorable boundary for empirical middle-correlation evaluation. Exploring alternative boundaries may involve multi-bit output masks, whose numerous possible combinations make empirical evaluation expensive. We propose a spectral DL search framework for efficiently evaluating multiple middle output masks. We use Hamming-weight-one correlations to select low-dimensional output-mask subspaces and apply the fast Walsh--Hadamard transform to jointly evaluate all masks in each selected subspace using shared samples. Experiments on 32-bit ciphers show that HW1-guided subspaces can recover DL candidates with correlation estimates comparable to those obtained through full-state spectral evaluation. We integrate this approach with round-split selection, automated outer-trail search, and independent fixed-key evaluation of the retained middle correlations. Experiments on SIMON, SIMECK, SPECK, and CHAM yield DL distinguishers with improved round coverage or estimated correlations. In particular, we obtain a 36-round SIMON128 distinguisher under single-trail outer-component evaluation and a 39-round result through refined outer-component evaluation. Our results further show that evaluating multi-bit masks at alternative middle-output boundaries can reveal more readily detectable correlations or reduce reliance on single-trail approximations.
Metadata
- Available format(s)
-
PDF
- Category
- Attacks and cryptanalysis
- Publication info
- Preprint.
- Keywords
- Differential-linear cryptanalysisWalsh-Hadamard transformLow-dimensional mask subspaces
- Contact author(s)
-
clzheng @ stu xidian edu cn
liuzhang academic @ gmail com
zlwang @ xidian edu cn - History
- 2026-11-22: approved
- 2026-11-19: received
- See all versions
- Short URL
- https://ia.cr/2026/2942
- License
-
CC BY
BibTeX
@misc{cryptoeprint:2026/2942,
author = {Chenlu Zheng and Liu Zhang and Zilong Wang},
title = {Differential–Linear Distinguisher Search via Walsh Spectra over Low-Dimensional Mask Subspaces},
howpublished = {Cryptology {ePrint} Archive, Paper 2026/2942},
year = {2026},
url = {https://eprint.iacr.org/2026/2942}
}