Re: [Ntp] Switching NTS to AES-GCM-SIV?

[Miroslav Lichvar <[e-mail]>]

On Tue, Apr 23, 2019 at 12:01:38PM -0400, Daniel Franke wrote:
> AES-GCM-SIV's misuse-resistance properties are weaker than AES-SIV's:
> its security bounds gradually degrade if you reuse the same nonce too
> many times. In mainstream use cases this isn't any cause for concern
> because nonce reuse should be rare, but I don't want to encumber our
> choice of MTI algorithm with the assumption that it will *never* be a
> concern.

In what way is NTS different from the mainstream in expected nonce reuse?

> With that said, in all but weird exceptional cases that may or may not
> exist, I definitely encourage implementers to support AES-GCM-SIV in
> addition to AES-SIV, and to put AES-GCM-SIV ahead of AES-SIV in their
> preference ordering when negotiating what to use.

I'm not sure I follow here. If you recommend implementations to
support and prefer AES-GCM-SIV, and it is found to be problematic in
the NTS use case later, what difference will it make that they already
support AES-SIV-CMAC? They will still need to be updated/reconfigured
to actually use it.

NTS implementations supporting both will just be more complex and have
a larger attack surface.

-- 
Miroslav Lichvar