AISchneier on Securityabout 1 hour ago
Measuring LLMs' Ability to Perform Cryptanalysis
2 min read
CryptanalysisBench measures LLM ability to find mathematical cryptanalytic attacks across 191 tasks. Five frontier models broke 65-86% of Tier 1 schemes. Anthropic's Mythos Preview discovered new vulnerabilities in Hawk and reduced-round AES.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
