Support age via PIV

[orolhawion]

I am not sure if this the right repository for this request, however, please feel free to move it wherever it suits best.

As of version 1.3.0 age supports post-quantum cryptography for encryption. In combination with age-plugin-yubikey you can use yubikeys to generate keys on the yubikey, in retired PIV slots if I remember correctly. The plugin is not yet PQ ready but it is on its way to get there.

Could you think of creating an age-plugin-nitrokey which basically does the same thing with nitrokeys (nk3)?

[CertainLach]

For my nitrokey 3 I have implemented age support using its openpgp card app here: https://github.com/CertainLach/age-plugin-gpg
It directly uses gpg-agent, making it possible to use it with gpg ssh authentication and other things cooperatively.

[orolhawion]

Nice, I suppose this is bound to the key algorithms that come with the gpg applet on the device in which case post quantum is still out, isn't it?

However, could you think of making it available through homebrew for Mac also?

[CertainLach]

Nice, I suppose this is bound to the key algorithms that come with the gpg applet on the device in which case post quantum is still out, isn't it?

Yes, some pq algorithm first needs to be implemented on gpg applet side.

However, could you think of making it available through homebrew for Mac also?

I don't have MacOS/don't use homebrew, but it should probably work on MacOS OOTB using Nix, as Nix supports it.

[sosthene-nitrokey]

Yes, some pq algorithm first needs to be implemented on gpg applet side

There are many challenges to solve before we can add PQ to the gpg applet, the main one being the modernisation of the standard, 4.0 of the OpenPGP smartcard specification is still WIP, and then there are issues with deploying PQ in a smartcard today. The secure element we use does not support any asymmetric PQ algorithm, so it would have to be a pure software implementation.

PQ in PIV has similar limitation, the standard needs define standard algorithm IDs before we could start working on PQ.

Note that currently OpenPGP supports many more algorithm than PIV.

[orolhawion]

The secure element we use does not support any asymmetric PQ algorithm, so it would have to be a pure software implementation.

No hardware support means that it is slow, right? Are you planning any new devices with PQ-ready hardware?

[sosthene-nitrokey]

Not sure if it will really be slower, like it's the case for RSA. PQ algorithm are not that slow even in software, so the constraint they have is more about key size/memory usage (which is a limitation for a software implementation where we're already constrained with regards to memory usage in the NK3).

The main risks with a software implementation are side channel attacks, as without dedicated hardware some attacks are much harder to prevent.

[orolhawion]

Does that mean that NK3 should be considered ineligible when PQ is a desired feature? and again, are you planning new devices with dedicated hardware for PQ?

[sosthene-nitrokey]

We are experimenting with PQ but we still don't have a roadmap to share on when we will be able to deploy PQ and with which product. As I said it still largely depends on the standards we implement being updated for PQ.