Improved (Related-Key) Differential-Based Neural Distinguishers for SIMON and SIMECK Block Ciphers

Abstract In CRYPTO 2019, Gohr made a pioneering attempt and successfully applied deep learning to the differential cryptanalysis against NSA block cipher Speck 32/64, achieving higher accuracy than pure distinguishers. By its very nature, mining effective features in data plays crucial role data-driven learning. this paper, addition considering integrity of information from training ciphertext pair, domain knowledge about structure is also considered into process improve performance. Meanwhile, taking performance differential-neural distinguisher Simon 32/64 as an entry point, we investigate impact input difference on hybrid distinguishers choose proper difference. Eventually, neural 64/128, Simeck 64/128. We obtain related-key differential-based round-reduced versions 64/128 for first time.