[est31]

The US federal government once expended 10 percent of the US's electric energy supply in the Manhattan project for getting weapons grade nuclear material. This gives a rough ballpark for the amount of energy they are willing to invest into major strategic advancements.

However, if you apply Landauer's principle, current factoring algorithms would require enough energy to boil all oceans on the earth, that's a lot even compared to the US's energy supply.

So algorithmic improvements are the real danger basically. Even if we discovered a decryption method now, and immediately everyone stopped using RSA, there would still be an immense impact because all the past encrypted traffic that someone might have stored somewhere suddenly becomes decryptable. And usually, traffic from 20 years ago is still relevant today.

[ameliaquining]

> This gives a rough ballpark for the amount of energy they are willing to invest into major strategic advancements.

I'd say, rather, that it gives a rough ballpark of how much the government was willing to invest in the 1940s, at the peak of its ability and willingness to take on projects of incredible scope. There was still a fair amount of this for a few decades after that, but not since the 70s. See https://rationalconspiracy.com/2012/06/03/why-doesnt-our-gov... for one person's take on this (though you don't have to go as far as she does, I think, to establish that applying Manhattan Project numbers to anything going on today will result in an overestimate).

[0xb100db1ade]

> current factoring algorithms would require enough energy to boil all oceans on the earth, that's a lot even compared to the US's energy supply.

Interesting. For what algorithm & key size?

I'd love to quote this. I've heard it before but I don't remember the source.

[est31]

https://eprint.iacr.org/2013/635.pdf

> Boiling all water on the planet (including all starfish) amounts to about 2^24 lakes of Geneva and leads to global security: 114-bit symmetric cryptosystems, 228-bit cryptographic hashes, 2380-bit RSA. This needs to be done 16 thousand times to break AES-128, SHA-256, or 3064-bit RSA.

I think this paper isn't using Landauer's bounds though, but conventional computers. So maybe my claim was wrong, because we aren't 16 thousand times away from Landauer's bounds but millions [1].

[1]: https://web.archive.org/web/20141219043239/http://www.bloomf...

[johndough]

But why the starfish.

[alfiedotwtf]

Because if you're already boiling the oceans, you probably won't have enough left over in the budget for marine conservation.