[Cfrg] PAKE review

[Brian Warner <[e-mail]>]

Dear CFRG,

Nick Sullivan asked me to participate in the PAKE selection process. My
specific use case is a tool/protocol I've developed named "Magic
Wormhole" (https://github.com/warner/magic-wormhole/), which uses PAKE
to let two humans introduce their computers to each other, for the
purpose of transferring a file, text message, or other secrets. I'm
encouraging its use as a provisioning protocol, to connect two
applications without the use of long-term human-memorized passwords. For
example, two users of a secure messenger could use it to safely exchange
their public keys and other address-book data by just speaking and
transcribing a short single-use phrase.

The magic-wormhole protocol currently uses SPAKE2 (a symmetric PAKE) on
an Ed25519 group, with an additional form of symmetry: the two sides do
not have to decide ahead of time which role they are playing. In SPAKE2
terms, the "M" and "N" elements are equal, which saves a roundtrip. This
program was inspired by work I did on Firefox Sync, back when it used
J-PAKE pairing codes (around 2009) to connect two browsers owned by the
same user.

Nick suggested that I might be able to help evaluate the four symmetric
PAKE nominations, both for their applicability to other contexts, but
particularly for magic-wormhole and other "provisioning" -style
applications, in which the codes are randomly generated and single-use.
I think augmented PAKEs are great, but they aren't as useful for these
kinds of protocols.

I'm happy to help out any way I can.

cheers,
 -Brian