Vaccari's Code

Claude reveals origins of ECDSA curves

10/18/2026

Cryptography is the backbone of our digital security, protecting everything from emails to financial transactions. For years, however, an enigma hung over one of the most widespread standards in this universe: ECDSA (Elliptic Curve Digital Signature Algorithm). The "seeds" that define its curves were provided without explanation by the NSA (National Security Agency) decades ago, generating distrust and speculation about possible "backdoors" – intentional secret access points that could compromise security. Now, this mystery has finally been unraveled, and the solution came from an unexpected source: an artificial intelligence model.

The Mystery of Cryptographic "Seeds"

ECDSA is the most widely used elliptic curve cryptography standard, adopted by NIST (National Institute of Standards and Technology) in 2000 as part of the FIPS 186-2 Digital Signature Standard. This standard lists a set of elliptic curves of different sizes and security levels. For curves over prime and binary fields, the specifications published a "seed" parameter whose hash value is used to define the curve's b coefficient.

The origin of these "seeds" has always been an open question. For example, the "seed" for the NIST P-256 curve was C49D3608 86E70493 6A6678E1 139D26B7 819F7E90, with no information about its provenance. This lack of transparency fueled the cryptographic community with theories, some even humorous, such as the "seeds" being a hash of a phrase like "Give Bob and Jerry a raise". The search for this origin became a kind of amateur "treasure hunt," with rewards offered to anyone who could uncover the secret. Unfortunately, the people involved in the original generation of these "seeds" were no longer available or couldn't recall the details, leaving the mystery intact.

The central concern was the possibility that the "seeds" might have been chosen in a way that contained hidden vulnerabilities, allowing the NSA (or any entity that knew the "secret") to break digital signatures generated by these curves. This distrust, though without concrete proof, cast a shadow over one of the pillars of modern digital security.

Claude Unravels the Enigma

The breakthrough happened in October 2026 (according to the source, a date that suggests the research nature of the finding). An Anthropic employee, curiously, asked an internal research model, Claude, if it could figure out how these "seeds" were generated, without providing any clues. In less than half an hour, Claude discovered the origin of the "seeds" for the NIST P-192, P-224, P-256, P-384, and P-521 curves.

The truth is that the "seeds" were not the result of any hash algorithm. Instead, they were the output of the lrand48 pseudorandom number generator (PRNG – Pseudorandom Number Generator, an algorithm that produces sequences of numbers that appear random but are deterministic if you know the initial "seed"), from a common C library of the time, initialized with a Unix timestamp – a number representing the amount of seconds elapsed since January 1, 1970.

Claude's approach was ingenious. It anticipated how the software that generated the curves must have worked, presuming the use of a PRNG from 90s C libraries. These older PRNGs are "weak" enough that their output can be recognized. Claude developed a "scanner" that slid bit by bit through each published "seed", testing whether adjacent blocks were consecutive outputs of generators from that era.

This is how Claude noticed a pattern in some upper bits of certain "seeds": they were all zero. Written as five 32-bit words, four of the ten "seeds" had a 0 in the uppermost bit for all 20 words. This would happen by chance with a probability of 1/5,810. This was a signature of the lrand48 linear congruential generator, which returns only the upper 31 bits of a 48-bit state.

For the "seeds" of the prime curves, which kept fewer bits from each output, Claude had to recover the PRNG's internal state from these truncated bits. Recovering the state of a truncated linear congruential generator is a classic lattice problem (a mathematical problem involving finding points in a multidimensional grid), solved with the LLL (Lenstra–Lenstra–Lovász) basis reduction algorithm. Claude tested various configurations and found a consistent pattern in five NIST "seeds" and five others from the ANSI X9.62 standard: lrand48() >> 23. Each recovered 48-bit state regenerated all 160 bits of its "seed", which does not happen by chance.

Why This Matters to Us (Builders)

This discovery is not a cryptographic flaw, a new attack, or a weakness in ECDSA curves. On the contrary, it increases our confidence in these curves. For many years, people suspected the origins of the NIST curve parameters. They were presented to NIST without any mention of how they were chosen, leading to speculation about intentional "backdoors".

Claude's findings make this hypothesis far less plausible. The "seeds" were generated in a very simple way: a common pseudorandom number generator, initialized from the system clock – something anyone can reproduce now. This leaves much less room for subversion and gives us confidence that the parameters are secure. There is no action to be taken or anything to be done differently; most researchers already trusted the curves.

For us, builders and technology decision-makers, this is an important lesson on several levels:

  1. Reinforced Cryptographic Confidence: It resolves a decades-old mystery and solidifies the foundation of a critical standard. Knowing that the "seeds" were generated in such a mundane way, without apparent malicious intent, is a relief for security architects and developers.
  2. The Power of AI as an Analytical Tool: It demonstrates the ability of advanced AI models not just to generate content or code, but to act as powerful analytical tools, capable of "reverse engineering" and finding patterns in complex and legacy systems where human intuition or traditional methods have failed for years.
  3. The Importance of Transparency: Although the origin was simple, the initial lack of transparency generated distrust. This reinforces the need for clear documentation and justifications for parameters in critical standards, even if the intention is benign.

Ultimately, the "mystery" of the cryptographic "seeds" was, in fact, a simple technological solution from the 90s, hidden in plain sight. AI helped us close this chapter, strengthening our confidence in the digital foundations we build.


Sources:


📬 Enjoyed this? Subscribe to the Vaccari's Code newsletter for the next wave of software & AI trends, straight to your inbox: Subscribe here

← all posts · listen to the episode →