Science & Tech · Cybersecurity · 2 hrs ago

DeepMind AI agents find faster ways to test weaknesses in old RSA encryption

DeepMind AI agents find faster ways to test weaknesses in old RSA encryption

Google DeepMind researchers used supervised AI agents to study weaknesses in outdated RSA-1024 encryption.

For one type of attack on 1024-bit keys, their method was about 1,600 times faster than the previous best approach.

The researchers also found that padding, which is added to digital signatures to help prevent forgery, offers little extra protection if an attacker controls about half of the signed data.

They tested their approach in a sandbox against an RSA-1024 benchmark that had already been solved in earlier academic work.

A separate method for analyzing multiple keys at once reduced the average cost of factoring each additional key by about 30 times compared with previous batch techniques.

RSA-1024 is an older standard, largely replaced by keys of 2048 bits or more.

The researchers say modern encryption remains secure, and argue that AI could help experts test future systems for weaknesses before attackers can exploit them.

1 source

Related