DeepMind's AI agents find two weaknesses in legacy 1024-bit RSA
Google DeepMind says human-guided AI agents cut the theoretical cost of forging some RSA signatures by about 1,600x. Modern 2048-bit keys are not affected.
- Signature forgery: ~1,600x more efficient (theory)
- Batch factoring: ~30x cheaper per extra key
- RSA-2048: not affected
What DeepMind found
Google DeepMind announced on 6 November that human-guided AI agents helped find two weaknesses in legacy 1024-bit RSA. The first targets signatures made over short fingerprints and is estimated to be about 1,600 times more efficient at forgery. The second lowers the marginal cost of each extra key in a large batch factoring attack by about 30 times.
The signature gain is a theoretical estimate for 1024-bit keys and was not run at full size. The paper is titled 'Forging 1024-bit RSA signatures in nearly SNFS time' and is on IACR.
Should you worry?
DeepMind's Pushmeet Kohli said the results only affect obsolete 1024-bit systems, and that modern protocols, which almost universally use 2048-bit or larger keys, remain secure. Security teams are still being urged to audit any leftover 1024-bit keys. Figures here come from press coverage of the announcement and have not been checked against the paper.
