Will any work mainly done by a Large Language model crack a widely used cryptography cipher in 2024. Could be for encryption, hashing. Could be symmetrical or assymetrical.
I mean a breakthrough that makes cryptanalysis of that cipher much easier, not that it reproduces a key that was in the training data that lets you decrypt a particular known set of data.
Differently stated, will an LLM make something like AES-192 significantly less secure.
Can be any widely used cipher, AES, blowfish, RSA. AES-192 would count.
Resolves based on my subjective interpretation of the market, not on any technicality. Or alternatively stated resolves based on spirit of the question, not on a strict literal interpretation of the above text.
@a2bb I don't have a watertight definition of the question that anticipates 99% of the likely cases, so I'm stating that I'll not resolve it as a literalist criteria, but the spirit of the question.
@a2bb I'd count AES-192, I wouldn't laugh someone out of the room if they wanted to use that in the encryption options.
@WilliamEhlhardt DES does not count, that's on the list of shame if you'd try to use that anywhere nowadays. I'm looking for novel cryptanalysis, a cipher that before didn't have any practical techniques but now is not advisable to use.
@ArmandodiMatteo Yes, that. Has to be used as actual cryptography, like something a company would publish in their 'use this cryptography' guidelines.
@ArmandodiMatteo For the record, I think it's very unlikely that it will happen, but I'd like to aggregate information if it does.