New insights on linear cryptanalysis

Abstract

Linear cryptanalysis is one of the most important cryptanalytic tools against block ciphers, thus modern block ciphers are always deliberately devised to avoid good long linear characteristics so as to resist linear cryptanalysis and its extensions. Differential-linear cryptanalysis, a powerful extension of linear cryptanalysis, has drawn much attention due to its applicability even in certain case that there is no good long linear characteristic of block ciphers. To further refine differential-linear cryptanalysis, we investigate the correlation distribution of differential-linear hull over random permutation and derive a concrete and concise correlation distribution accordingly. Theoretically, this could make differential-linear cryptanalysis more reasonable and precise. Moreover, the newly-proposed correlation distribution could lead to an interesting potential for improving the effectiveness of differential-linear cryptanalysis.