Comment Re:AES crypto attacks (Score 1) 24
The attack is against 7-round AES. As I read the abstract, they claim to optimise an attack from 2013.
Cryptanalysis of 7-Round AES via the Algebraic Structure of its S-box
Milad Nasr and Nicholas Carlini
The AES S-Box is not a random permutation; we show that its algebraic structure allows for improved attacks on 7 rounds in the single-key setting. We focus
our efforts on extending a 2105 chosen plaintext and 299-time algorithm designed by Derbez, Fouque, and Jean [DFJ13]. Our core contribution is an algebraic Möbius bridge that exploits the invert-then-affine-transform structure of the AES S-box to eliminate one of the nine guessed key bytes required by [DFJ13]. This immediately leads to a factor-of-256 reduction in the number of key guesses required. But calculating the necessary Möbius Transform is not free, and the work required to perform the naïve calculation cancels the savings. We therefore present several improvements that reduce the time required to calculate this transform; depending on the exact methods used for accounting, we reduce the attack’s runtime to between 289.3 and 291.4, at the same data complexity. This paper presents these improvements along with the computational techniques we used to help us empirically gain confidence in the correctness of our findings, something that was particularly important because these improvements were discovered autonomously via a large language model fully end to end. We conclude by remarking on the future of LLM-assisted research in cryptography and computer security more generally.
