Claude reveals ECDSA curve origins

934 points · 340 comments on HN · read original →

Points and comments are a snapshot, not live.

Claude reverses the 1990s PRNG behind NIST's ECDSA curve seeds, ending a 25-year mystery.

Anthropic employees applied Claude to a long-standing cryptographic puzzle: the seeds used to define NIST's P-192, P-224, P-256, P-384, and P-521 ECDSA curves, provided by the NSA without explanation. Claude identified the output of the C library's lrand48 generator, seeded with Unix timestamps between 1997 and 1998. The seeds are not hashes of any phrase, as previously rumored. Claude found some base points (P-192 and five ANSI X9.62 example curves) came from FreeLIP's random number generator. The P-256 seed was candidate 1842 of its run. The B-163 binary-curve seed remains unexplained. The $36,864 bounty will be donated to the Internet Security Research Group. The results are reproducible with provided code.

What commenters are saying

The thread's dominant reaction is relief that the long-standing conspiracy theory is resolved, though many note that cryptographers had already considered the backdoor concern far-fetched. Several commenters express skepticism that NIST would have burned trust by failing to disclose the simple generation method. A secondary concern arises: that Anthropic (and other labs) gate access to such capabilities behind safety filters, widening a moat between the public and frontier models. One commenter reports being blocked for trivial cryptography queries, while Anthropic employees had access to Claude's full reasoning.