DeepMind AI Agents Find Weaknesses in 1024-bit RSA
Google DeepMind said human-guided AI agents uncovered two weaknesses in legacy 1024-bit RSA systems, improving attacks involving signature forgery and batch integer factorization. Zhidx reported that two agents reproduced a prior roughly five-month 1024-bit signature-forgery computation in 3 days and 22 hours, reducing CPU usage from about 1,380 to 391 core-years. The agents wrote about 33,500 lines of code and completed 336 distributed computing experiments while human researchers set the goals and conditions and reviewed the results.
The work shows AI agents assisting with program development and cryptanalysis computation while human researchers define the research conditions and review results, according to Google DeepMind and Zhidx.
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ https://x.com/pushmeet/status/2118752095859794390
@pushmeet: https://x.com/i/article/2118746079923437568 https://x.com/pushmeet/status/2118752095859794390
#Google DeepMind#University of California San Diego#Inria#Fan Factories