Governance, Risk and Compliance, Critical Infrastructure Security

AI model reconstructs origins of widely used security standard numbers

Innovative ai showcase global tech summit 2025 digital innovations futuristic environment interactive viewpoint advancements in ai technology

As reported by Cyber Insider. An AI model named Claude has successfully reconstructed the origins of numerical seeds used in widely adopted security standards, resolving a long-standing mystery about their creation.

Anthropic's Claude AI identified that the seeds for five standardized mathematical curves, which underpin the Elliptic Curve Digital Signature Algorithm (ECDSA), were generated using a conventional random-number generator. This generator was initialized with dates and times from the 1990s, specifically between May 1997 and April 1998. This finding addresses decades of speculation that the U.S. National Security Agency (NSA) might have deliberately introduced hidden weaknesses into these widely used security standards. The ECDSA is crucial for verifying authenticity and detecting tampering in digital communications. While the NSA provided the seeds, their origin remained unexplained until Anthropic's Claude model, using mathematical techniques to reconstruct the generator's internal state, pinpointed the use of the lrand48 generator. Although the exact reason for skipping certain candidate curves remains unclear, the reconstruction of the seeds' origins does not compromise private keys or enable forged signatures, thereby strengthening confidence in the security standards.

Source: Cyber Insider

Get daily email updates

SC Media's daily must-read of the most current and pressing daily news

By clicking the Subscribe button below, you agree to SC Media Terms of Use and Privacy Policy.

You can skip this ad in 5 seconds