search

jedisct1 / libsodium

A modern, portable, easy to use crypto library.
https://libsodium.org
Other
12.06k stars 1.72k forks source link

RFC9381 ECVRF implementation #1188

Closed iquerejeta closed 1 month ago

iquerejeta commented 2 years ago

Implementation of RFC9381.

Made a change to ed25519_ref10 for performance reasons. Mainly, I have included a variable base variable time scalar multiplication (to compute a * U + b * V for variable U and V).

Implements #1177

jedisct1 commented 10 months ago

https://www.rfc-editor.org/rfc/rfc9381.pdf

iquerejeta commented 10 months ago

@jedisct1 , does this mean that we are good to push this forward? If yes, I'll make a pass to make sure that the implemented version is what got finalised in the draft, and mark it as ready.

jedisct1 commented 10 months ago

@iquerejeta Yes :)

iquerejeta commented 10 months ago

Got some trouble reproducing the undefined reference to 'crypto_core_ed25519_scalar_negate' and undefined reference to 'crypto_core_ed25519_scalar_negate' errors locally. I made sure I was configuring with the same call as the CI action: 

env CPPFLAGS="-DDEV_MODE=1" ./configure --disable-dependency-tracking --enable-minimal

But couldn't reproduce it.

This is the failing action https://github.com/jedisct1/libsodium/actions/runs/6095043654

iquerejeta commented 10 months ago

So, I've managed to reproduce them, but no progress in debugging. I'll try again tomorrow. Any idea why we might have 

Undefined symbols for architecture arm64:
...
ld: symbol(s) not found for architecture arm64

type errors for symbols that are exposed in the public API when called within the library?

jedisct1 commented 10 months ago

Errrrr... this is super weird.

How did you manage to reproduce it? Is it non-deterministic?

jedisct1 commented 10 months ago

Actually it's not weird.

The core_ed25519 functions are not available in MINIMAL builds.

iquerejeta commented 9 months ago

Ok, makes sense. What is your preferred way forward? To have core_ed25519 exposed in the MINIMAL build, or instead have from_string and scalar_negate as part of ed25519_ref10.c? Seeing how the library is organised, I presume the second, but just want to check with you before making further changes.

jedisct1 commented 9 months ago

Yeah, I agree that the second option looks better.