Cryptology ePrint Archive: Report 2021/310

A New Neural Distinguisher Considering Features Derived from Multiple Ciphertext Pairs

Yi Chen and Yantian Shen and Hongbo Yu and Sitong Yuan

Abstract: Neural aided cryptanalysis is a challenging topic, in which the neural distinguisher (ND) is a core module. In this paper, we propose a new ND considering multiple ciphertext pairs simultaneously. To our best knowledge, this is the only ND except for the ND proposed by Gohr at CRYPTO’19. Taking Gohr’s ND as the strong baseline model, we perform an in-depth analysis of our new ND. First, applications to five different ciphers show that our NDs achieve higher distinguishing accuracy. Second, we prove that our ND successfully captures features derived from multiple ciphertext pairs. Third, we further show how to perform various key recovery attacks with this new ND. More advantages of our ND are further discovered in key recovery attacks. Taking the neural aided statistical attack (NASA) as an example, we prove that the data complexity can be reduced by replacing Gohr’s ND with our ND.

Category / Keywords: secret-key cryptography / Cryptanalysis, Neural distinguisher, Differential cryptanalysis, Deep learning, Data reuse

Date: received 8 Mar 2021, last revised 16 Aug 2021

Contact author: chenyi19 at mails tsinghua edu cn

Available format(s): PDF | BibTeX Citation

Version: 20210816:100908 (All versions of this report)

Short URL: ia.cr/2021/310


[ Cryptology ePrint archive ]