Anthropic researchers (John Wood, Henry de Valence, Ava Howell, Zack Voase, Lucy Sweet, Amir Omidi, Paul Crowley, Iain McGinniss, Nicholas Carlini, and Steve Weis) claimed a bounty on October 15, 2026 by showing the NIST elliptic-curve seeds were generated with the POSIX drand48 RNG seeded with UNIX timestamps in 1997-98. They donated the prize to ISRG, bringing the total to $36,864. Reimplementing the generation in Go confirmed that srand48(timestamp) produced byte sequences which, when tested under the ANSI X9.62 seed-to-curve process, yield P-192, P-224, P-256, P-384, and P-521 (and prime239 variants) at particular timestamps. The published mapping of timestamps to accepted seed indices and several nonpublished accepted seeds corroborates the claim and is documented in Anthropic’s writeup.
Two undocumented selection rules remain: seeds were rejected if the bit length of the curve order n exceeded the bit length of the base field p (a looser check than n < p), and a second unexplained pattern of rejections halves the expected published set, possibly due to RNG consumption or legacy bugs. Security implications are decisive: the timestamp+seed space gives at most ~30 bits of adversarial freedom, eliminating plausible NOBUS/backdoor scenarios and refuting Bernstein-Lange targeting concerns. Combined with modern ECDSA fixes and verified implementations, this restores practical confidence in NIST curves even as post-quantum signatures approach deployment.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.