Paper 2024/1998 Original: eprint.iacr.org/2024/1998 ↗

Impossible Differential Automation: Model Generation and New Techniques

Emanuele Bellini
Paul Huynh
David Gerault
Andrea Visconti
Alessandro De Piccoli
Simone Pelizzola
Abstract

In this paper, we aim to enhance and automate advanced techniques for impossible differential attacks. To demonstrate these advancements, we present improved attacks on the LBlock and HIGHT block ciphers. More precisely, we (a) introduce a methodology to automatically invert symmetric ciphers when represented as directed acyclic graphs, a fundamental step in the search for impossible differential trails and in key recovery techniques; (b) automate the search for impossible differential distinguishers, reproducing recent techniques and results; (c) present a new hybrid model combining cell-wise properties and bit-wise granularity; (d) integrate these techniques in the automated tool CLAASP; (e) demonstrate the effectiveness of the tool by reproducing a state-of-the-art 16-round impossible differential for LBlock previously obtained using a different technique and exhibiting a new 18-round improbable trail; (f) improve the state-of-the-art single-key recovery of HIGHT for 27 rounds, by automating the use of hash tables to current state-of-the-art results.

Metadata
Available format(s)
PDF
Category
Secret-key cryptography
Publication info
Preprint.
History
2025-01-28: last updated
2024-12-11: published
Short URL
https://ia.cr/2024/1998
License
https://creativecommons.org/publicdomain/zero/1.0/

BibTeX

@misc{cryptoeprint:2024/1998,
      author = {Emanuele Bellini and Paul Huynh and David Gerault and Andrea Visconti and Alessandro De Piccoli and Simone Pelizzola},
      title = {Impossible Differential Automation: Model Generation and New Techniques},
      howpublished = {Cryptology {ePrint} Archive, Paper 2024/1998},
      year = {2024},
      url = {https://eprint.iacr.org/2024/1998}
}
Note: In order to protect the privacy of readers, this archive does not use cookies or embedded third party content.