Paper 2022/1467

A Cipher-Agnostic Neural Training Pipeline with Automated Finding of Good Input Differences

Abstract

Neural cryptanalysis is the study of cryptographic primitives throughmachine learning techniques. Following Gohr’s seminal paper at CRYPTO 2019, afocus has been placed on improving the accuracy of such distinguishers against specific primitives, using dedicated training schemes, in order to obtain better key recovery attacks based on machine learning. These distinguishers are highly specialized and not trivially applicable to other primitives. In this paper, we focus on the opposite problem: building a generic pipeline for neural cryptanalysis. Our tool is composed of two parts. The first part is an evolutionary algorithm for the search of good input differences for neural distinguishers. The second part is DBitNet, a neuraldistinguisher architecture agnostic to the structure of the cipher. We show thatthis fully automated pipeline is competitive with a highly specialized approach, inparticular for SPECK32, and SIMON32. We provide new neural distinguishers forseveral primitives (XTEA, LEA, HIGHT, SIMON128, SPECK128) and improve overthe state-of-the-art for PRESENT, KATAN, TEA and GIMLI.