Re: [Cfrg] Help with the use of contexts

[Adam Langley <[e-mail]>]

On Wed, Jan 4, 2017 at 1:55 PM, Sean Turner <[e-mail]> wrote:

> The TLS WG is nearing the end of our journey moving EC-based algorithms
> for TLS 1.2 (and earlier) from Informational to Standards track [0].  While
> we were doing that we also added in 25519 and x448 as well as EdDSA.
>
> I’d like to get some input from the CFRG on the use of contexts; the
> "context label" is a way to provide domain separation between signatures
> made in different contexts, avoiding cross-protocol attacks.  s10.3 of
> draft-irtf-cfrg-eddsa includes the following:
>
> Contexts SHOULD NOT be used opportunistically, as that kind of use
> is very error-prone.  If contexts are used, one SHOULD require all
> signature schemes available for use in that purpose support
> contexts.
>
> This is great advice for new protocols because it’s easy to make all the
> schemes the same, but for existing protocols like TLS where there’s zero
> chance of obsoleting the existing signature schemes and defining new
> signature schemes with contexts it makes you wonder what
> “opportunistically” means. I.e., would setting a context parameter for
> Ed448 and no other already defined signature scheme be considered
> opportunistic?
>

Domain separation for signed values is important and TLS already defines
context strings for signature operations:
https://tools.ietf.org/html/draft-ietf-tls-tls13-18#section-4.4.2

The way that this is constructed (due to me) is generic for any signature
scheme. (Basically just have the context string be NUL-terminated at the
beginning of the signed message.)

So for TLS I believe that the issue is already taken care of. It would also
probably just add pain for implementers if the context string were to be
duplicated as an input to the signature scheme where signature schemes
support it.


Cheers

AGL

-- 
Adam Langley [e-mail] https://www.imperialviolet.org