Title: AI for the Cryptanalysis of Post-Quantum Algorithms

Abstract: With the push toward post-quantum cryptography (PQC), assessing the security of new cryptographic schemes has never been more important. In this talk, I’ll discuss how machine learning—specifically language models—can be used to attack lattice-based and code-based cryptosystems. I’ll go over recent work on using AI to break Learning With Errors (LWE) and its variants, including structured versions like Module LWE used in Kyber. I’ll also touch on early results applying similar techniques to code-based cryptography, such as McEliece and BIKE. The goal is to explore what AI can (and can’t) do for cryptanalysis and what this means for the security of post-quantum cryptography.