摘要
1 Introduction In CRYPTO 2019,Gohr[1]innovatively integrated deep learning with differential cryptanalysis,specifically applied to SPECK32/64,resulting in the development of a neural distinguisher(ND)that outperforms the DDT-based distinguisher(DD).Subsequently,a hybrid distinguisher(HD)was introduced,combining the strengths of ND and a classical differential(CD)and the practical realization of 11-and 12-round key recovery attacks is launched.In 2022,Lyu et al.[2]further enhanced Gohr's framework,applying it to SIMECK32/64.To more deeply evaluate the security of SIMECK32/64,we made some improvements for differentialneural cryptanalysis,as listed below.
基金
supported by the National Natural Science Foundation of China(Grant Nos.62172319,62172427)
the Fundamental Research Funds for the Central Universities(No.QTZX23090)
the Postgraduate Scientific Research Innovation Project of Hunan Province(No.CX20220016).
