Cloudflare certificate changes

lionel wrote:

thus “old” OS/browser that do not trust the Let’s Encrypt root will not recognise new Let’s Encrypt certificates (that do not “chain back” to IdenTrust) as valid.

Ok, thanks for that precision. I suppose it concerns very old browser/OS…

Peter wrote:

Off topic I know but this is such a highly specialised thread anyway

Indeed, but I thought it was the cause, because embedded cryptography of products (either hw or sw) must be 10y in advance, that’s the reason behind my thought.
Nvidia is largely promoting they computing power for AI, with which they will gather a lot of money in the coming years, but their products can be used for RSA (and I suppose ECDSA).

Peter wrote:

AFAIK most RSA stuff is now 4096 bits

Oh no, not by far. The certificate used by EuroGA is 2048 bits RSA.

Peter wrote:

so incredibly hard to crack by any known method?

It is hard to crack by any known method on any computer known to have been manufactured. RSA with 4096 bit keys is not hard to crack if one were to build a quantum computer with 8192 qubits. Likewise, RSA with 2048 bit keys is not hard to crack if one were to build a quantum computer with 4096 qubits.

For comparison, there is now a quantum computer with 1180 qubits, but I think I remember I saw an explanation that this is “noisy qubits” or limited in some way, and not “ideal theoretical qubits” which is what the necessary qubits in the previous paragraph are. I’m not sure if the best current publicly known quantum computer has about 100 perfect qubits or about 400.

lionel wrote:

I think I remember I saw an explanation that this is “noisy qubits” or limited in some way, and not “ideal theoretical qubits” which is what the necessary qubits in the previous paragraph are. I’m not sure if the best current publicly known quantum computer has about 100 perfect qubits or about 400.

No quantum computer can have perfect qubits – they’re all “noisy” – they have a nonzero likelihood of spontaneously lose coherence, i.e. revert to a classical bit. The issue is that the more qubits you have the greater the chance that at least one of them will lose coherence, thus breaking your computation. On the other hand, the more qubits you have the more error correction you can do due to redundancy. So the question is how many “effective” qubits the computer has. Quantum computing is not my speciality, but from what I read in Wikipedia, you need “tens of thousands” of qubits to be able to break RSA.

Peter wrote:

Interesting… can these devices run any actual code?

That depends on what you mean by “running code”. They are kind of coprocessor.

I’d be extremely surprised should anyone report issue related to this.