Anthropic's Claude AI has solved a mystery that has puzzled cryptographers and security experts for a quarter of a century in half an hour. The model has revealed how the US National Security Agency (NSA) created key elements for global encryption standards in the 1990s.
Anthropic's Claude AI has solved a mystery that has puzzled cryptographers and security experts for a quarter of a century in half an hour. The model has revealed how the US National Security Agency (NSA) created key elements for global encryption standards in the 1990s.
Back in 2000, the US approved digital signature standards that are still used worldwide. They are based on so-called "elliptic curves" - mathematical formulas that protect our data on the Internet, writes Cyber Press.
However, the NSA used special initial digital codes (so-called seed values or seeds) to create them. For decades, cryptographers did not understand where the secret service got these numbers from. Such secrecy gave rise to fears and conspiracy theories: it seemed to many that the NSA had deliberately chosen such parameters in order to leave itself “secret doors” (bookmarks) for surveillance.
Some researchers even suggested that humorous phrases in English were encrypted in these numbers.
Instead of guessing encrypted phrases, Claude's artificial intelligence approached the task like a programmer from the 1990s.
He analyzed how common programming languages (C, in particular) worked at the time, and guessed that the NSA developers hadn't come up with anything complicated. They simply used a standard random number generator, which they ran with a timestamp (date and time).
Claude examined the digital codes and found that they were created between May 1997 and April 1998. For example, for the popular P-256 security standard, the required code was generated on August 16, 1997.
The study suggests that the NSA was likely not trying to manipulate mathematics, but simply using standard tools of its time. Rather, it is a reconstruction of events from 25 years ago that added transparency to global security standards.
For solving this mystery, Anthropic will receive a charitable reward of $36,864, which will be donated to the Internet Security Research Group, an organization that focuses on Internet security.
