Re: [kitten] draft-hansen-scram-sha256 and incorporating session hashing for channel binding

[Simon Josefsson <[e-mail]>]

Den Fri, 22 May 2015 20:16:15 -0400
skrev Re: [kitten] draft-hansen-scram-sha256 and incorporating
session         hashing for channel binding:

> On 3/2/15 2:23 PM, Alexey Melnikov wrote:
> >
> > On 25/02/2015 15:25, Simon Josefsson wrote:
> >> Tony Hansen <[e-mail]> writes:
> >>
> >>> On 2/24/15 6:56 AM, Stephen Farrell wrote:
> >>>> But in addition, there were two substantive issues that ought be
> >>>> resolved before IETF LC:
> >>>>
> >>>> 1. a new channel binding or requiring tls-session-hash (and I
> >>>> guess some explanatory text about why that is good/needed)
> >>> To recap:
> >>>
> >>> Simon Josefsson made this comment:
> >>>
> >>>> Since SCRAM was published, we have learned that the tls-unique
> >>>> channel binding is insecure -- it would be nice if we could
> >>>> combine the SHA256 update with another default channel binding
> >>>> type to resolve that problem.  In my view, the problem with
> >>>> SCRAM today isn't primarily its use of SHA1 but it's broken
> >>>> channel binding.
> >>> Martin Thompson responded:
> >>>
> >>>> We have a solution for that:
> >>>> https://tools.ietf.org/html/draft-ietf-tls-session-hash
> >>> I've read through tls-session-hash and am unsure how to proceed
> >>> here.
> >>>
> >>> One of my goals when proposing SCRAM-SHA-256 was to not change the
> >>> protocol at all, other than updating the hash algorithm.
> >>>
> >>> I'm not sure how to incorporate a recommendation for session
> >>> hashing here. I'm thinking this would be best handled by adding
> >>> something to the Security Considerations section. Does that seem
> >>> right?
> >>>
> >>> Would anyone be willing to suggest text changes for these
> >>> comments?
> >> I believe the problem is that RFC 5802 is insecure as currently
> >> specified, and we are bringing this up as a problem with your
> >> draft, which is unfair to you since the problem was not introduced
> >> by you.
> >>
> >> If people like the tls-session-hash approach (I'm not in that
> >> category, but there may be consensus around it), the proper fix is
> >> to update RFC 5802 and reference tls-session-hash as a normative
> >> reference.  This will take care of the problem, as you could copy
> >> that text into your document.  If you are looking for a text
> >> change here, it would be:
> >>
> >>    To be secure SCRAM-SHA-256-PLUS has to be used over a TLS
> >> channel that
> >>    MUST have [TLS-SESSION-HASH] negotiated.
> >>
> >> Personally, I would prefer to change to another mandatory channel
> >> binding that is secure for all TLS versions.
> > Before we make the decisions between referencing tls-session-hash
> > versa a new channel binding, can you sketch out how the new channel
> > binding is going to be defined?
> >
> > (If we choose to define a new channel binding, I think Kitten WG is
> > the right place for doing this work.)
> 
> I added the following to the security section:
> 
> >     The strength of this mechanism is dependent in part on the
> hash-iteration count, as denoted by "i" in
> >     <xref target="RFC5802"/>.
> >     As a rule of thumb, the hash-iteration count should be such
> > that a
> modern machine will take 0.1 seconds
> >     to perform the complete algorithm; however this is unlikely to
> > be
> practical on mobile devices and
> >     other relatively low-performance systems.
> >     At the time this was written, the rule of thumb gives around
> 15,000 iterations required;
> >     however an iteration count of 4096 takes around 0.5 seconds on
> current mobile handsets.
> >     This computational cost can be avoided by caching the ClientKey
> (assuming the Salt and iteration count is stable).
> >     Therefore the recommendation of this specification is that the
> iteration count SHOULD be at least 4096, but careful
> >     consideration ought to be given to using a significantly higher
> value, particularly where mobile use is less important.
> 
> If anyone has suggestions for additions to this, or changes, please
> suggest it now.

This text has nothing to do with the tls-session-hash issue?

If you don't want to ship a known insecure protocol, I believe the
options are to either include text similar to what I proposed above, or
to replace tls-unique with a secure TLS channel binding (e.g.,
draft-josefsson-sasl-tls-cb).  Shipping a known insecure protocol is
another option, but then you should add a security consideration
explaining that tls-unique is not secure without tls-session-hash and
that consistency with (the also insecure) RFC 5802 was deemed more
important security.

/Simon