Repository navigation
Replies: 1 comment 1 reply
|
Yes, this is in scope, and technically it should be feasible. RS-Key already has an ML-DSA implementation used by the experimental post-quantum FIDO2 support, so the cryptographic primitive itself is already there. The missing part is integrating it properly into the PIV applet and its key/certificate handling. SoloKeys currently has ML-DSA-44 support in PIV as well. However, as far as I understand it, their current implementation is necessarily a non-standard/vendor extension rather than an implementation of a finalized PIV PQC specification. NIST only published the initial working drafts for PQC support in PIV in June 2026. They cover ML-DSA and ML-KEM, but parts of the protocol are still being defined — including PIV algorithm identifiers, key references, command behavior and related data formats. Because of that, I would prefer not to introduce another incompatible vendor-specific PIV dialect into RS-Key and then have to break it once the standard is finalized. My current preference is therefore to follow the NIST work and implement ML-DSA for PIV once the specification becomes sufficiently stable. An experimental/vendor implementation before that is still possible, but I would keep it explicitly experimental. There is also a host-side interoperability issue: having ML-DSA on the token is only one part of the stack. PKCS#11/OpenSC, X.509 tooling and the CA software also need to understand the resulting keys and signatures. Step CA adding support would make this use case considerably more practical. |
Uh oh!
There was an error while loading. Please reload this page.
Thanks for this project! It's already great and very much appreciated :)
Not sure if this is in scope, but one feature I have been looking for quite some time is a small, local PKI with PQC.
I want to be able to have a root key/cert with 256-bit equivalent security and a lifetime of 5-10 years, where I don't need to worry about whether a quantum computer could break it. If it were solely for TLS, that would be of less concern, but a local PKI might also be used for other purposes. I don't want NIST curves if it can be avoided and RSA keys would need to be >4096 bits, so currently the "best" available option would have been Pico-HSM with Ed448, although, that of course is still not PQC. I attempted that route previously, but since their API is essentially closed off now it proved to be difficult, to say the least.
I don't need advanced HSM functionality such as many keys or backup mechanisms. In theory (I'm unsure whether it's feasible, but a draft for the NIST standard already exists), it would suffice if the PIV module supported the ML-DSA PQ algorithms. It should then work (soon) e.g. with Step CA (smallstep/certificates#2785). Intermediate keys/certs (at least in the TLS case) would still most likely need to use NIST curves such as P-384 for compatibility reasons, however.
Disclaimer: I'm aware that Let's Encrypt etc. exists, and I know about the limitations of the RP2350 hardware. I still think this would be a valid use case however, and an improvement over software-based PKI, even when using an offline signing device.
All reactions