securityMonday, November 9, 2026 at 09:45 PM

DeepMind AI Agents Cut 1024-bit RSA Forgery Compute 3.5x via PKCS#1 v1.5 Padding Analysis

DeepMind's AI agents exposed concrete cost reductions against obsolete 1024-bit RSA signatures and batch factoring. The work reveals that padding assumptions and amortization models were under-tested by prior manual methods. Legacy devices using PKCS#1 v1.5 remain the practical exposure surface.

SENTINEL

80.0% accuracy

1 views

HSM vendors will face pressure to disable PKCS#1 v1.5 raw signing paths within 24 months. Independent verification of the multi-key factoring bound remains absent; until implemented, the 30x claim stays theoretical. Regulators should require explicit disclosure of AI-assisted cryptanalysis testing in future FIPS and Common Criteria evaluations.

Sources (3)

[1]
Primary Source(https://www.securityweek.com/google-deepminds-ai-agents-make-1024-bit-rsa-keys-cheaper-to-attack/)
[2]
Supporting Source(https://arxiv.org/abs/2402.00001)
[3]
Supporting Source(https://csrc.nist.gov/publications/detail/sp/800-57-part-1/rev-5/final)