Question for app developers, like Enigmail etc. - Identicons

[Daniel Kahn Gillmor]

Hi Stefan--

I think you're asking about two sort of different things.

on the one hand, you're asserting that the 32-bit keyid isn't sufficient
for any sort of cryptographic verification.  that's absolutely correct,
and enigmail really shouldn't be exposing the 32-bit keyID to humans
where it can avoid doing so.  I've written more about this here:

  https://debian-administration.org/users/dkg/weblog/105

You're also asking about graphical representations of the cryptographic
identity -- a graphical representation of a fingerprint, basically.
The community has seen several different proposals of graphical
fingerprint representations in the past, and every one i've seen
gets stuck when faced with the hard questions.  In particular:

 * is the goal *recognition* of the fingerprint (i.e. "does this
   fingerprint look sufficiently similar to the one i've seen in the
   past for me to remember it?"), or is the goal *distinguishing* from a
   maliciously-crafted fingerprint (i.e. "am i certain that this
   fingerprint is an exact match of one that i expect to see from the
   peer who i think should have been signing this e-mail?")

 * In the "recognition" model, it's not clear that any
   cryptographically-strong guarantees are made to the user.  So why tie
   the visual identity to the cryptographic identity if we think it's
   spoofable?

 * in the "distinguishing" model, it's not clear that any of the schemes
   i've seen are actually better for most humans against a dedicated
   attacker who crafts fingerprints to make visual identities that look
   similar.  do you have any studies showing this capability against a
   motivated and technically capable attacker?

I'd generally think that if you're looking for a tool to help people
remember and recognize keys that they've seen before, then a mail user
agent is in a great position to do exactly that: just tell the user
explicitly what they've seen before, how often, etc.  why depend on the
human visual cortex or on human ability for numeric recall?

      --dkg
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 832 bytes
Desc: not available
URL: </pipermail/attachments/20170604/2a900399/attachment.sig>