On Impossible Boomerang Attacks Application to Simon and SKINNYee

• Published in: IACR Transactions on Symmetric Cryptology Vol. 2024; no. 2; pp. 222 - 253
• Main Authors: Bonnetain, Xavier, Cordero, Margarita, Lallemand, Virginie, Minier, Marine, Naya-Plasencia, María
• Format: Journal Article
• Language: English
• Published: Ruhr Universität Bochum 18.06.2024
• Subjects: Computer Science; Cryptography and Security; Cryptanalysis; Impossible boomerang attack; Simon; SKINNYee
• ISSN: 2519-173X; 2519-173X
• DOI: 10.46586/tosc.v2024.i2.222-253

The impossible boomerang attack, introduced in 2008 by Jiqiang Lu, is an extension of the impossible differential attack that relies on a boomerang distinguisher of probability 0 for discarding incorrect key guesses. In Lu’s work, the considered impossible boomerang distinguishers were built from 4 (different) probability-1 differentials that lead to 4 differences that do not sum to 0 in the middle, in a miss-in-the-middle way.In this article, we study the possibility of extending this notion by looking at finerlevel contradictions that derive from boomerang switch constraints. We start by discussing the case of quadratic Feistel ciphers and in particular of the Simon ciphers. We exploit their very specific boomerang constraints to enforce a contradiction that creates a new type of impossible boomerang distinguisher that we search with an SMT solver. We next switch to word-oriented ciphers and study how to leverage the Boomerang Connectivity Table contradictions. We apply this idea to SKINNYee, a recent tweakable block cipher proposed at Crypto 2022 and obtain a 21-round distinguisher.After detailing the process and the complexities of an impossible boomerang attack in the single (twea)key and related (twea)key model, we extend our distinguishers into attacks and present a 23-round impossible boomerang attack on Simon-32/64 (out of 32 rounds) and a 29-round impossible boomerang attack on SKINNYee (out of 56 rounds). To the best of our knowledge our analysis covers two more rounds than the (so far, only) other third-party analysis of SKINNYee that has been published to date.