Re: Message takeover attacks against S/MIME

On Mon, Mar 28, 2016 at 9:49 AM, Jim Schaad <[email protected]> wrote:

> I am open to suggestions, but until somebody gives be a better answer I do
> not know how to do the following:
>
> > > * A better mechanism than sign-encrypt-sign to get both authentication
> and
> > confidentiality (or at least a non-strippable way of doing it)
>
> Using and AEAD algorithm only solves part of the problem as one can do an
> encrypt-sign but the outer integrity cannot be validated by a third party.
> This is also set to be able for third parties to add the outer signature
> layer and that has been used in several specifications so it is not clear
> to me that this is one issue that needs to be fixed.
>
> I do to totally agree that we need to start moving to AEAD algorithms
> however.
>

(Caveat: I'm not familiar with these techniques, and am posting in part to
learn if this is a good idea.)

Have folks here considered authenticated key exchange to generate per
message symmetric key for AEAD?  Making the exchange asynchronous appears
to be possible as was done in TextSecure.  There more description in the
SoK paper (link <http://cacr.uwaterloo.ca/techreports/2015/cacr2015-02.pdf>)
around 13-16.  Criticism of TextSecure suggest though, the pre-keying could
cause problems for email applications.

-Wei


>
> jim
>
> > -----Original Message-----
> > From: smime [mailto:[email protected]] On Behalf Of Sean Turner
> > Sent: Monday, March 28, 2016 6:11 AM
> > To: Wei Chuang <[email protected]>
> > Cc: IETF SMIME <[email protected]>
> > Subject: Re: [smime] Message takeover attacks against S/MIME
> >
> > If the patient is on the table, then all of these look like items we’d
> need to
> > address.
> >
> > spt
> >
> > > On Mar 26, 2016, at 09:57, Wei Chuang <[email protected]> wrote:
> > >
> > > Some more work item suggestions (I'm posting on behalf of someone who
> > wishes to avoid the rough and tumble of discussion on the mailing list
> though
> > reads it):
> > > * Making ECC an official part of S/MIME, RFC5753 is only informational.
> > > * Authenticated encryption.
> > > * A better mechanism than sign-encrypt-sign to get both authentication
> and
> > confidentiality (or at least a non-strippable way of doing it)
> > > * Dropping obsolete ciphers and hashes from the spec: drop SHA-1, MD5,
> etc.
> > > * A modernized RFC 4134 (practical examples of S/MIME messages, it
> could
> > avoid the fiasco of 3rd party forgetting ASN-1 tags in their messages).
> > >
> > > -Wei
> > >
> > >
> > > On Sun, Mar 13, 2016 at 9:38 AM, Wei Chuang <[email protected]> wrote:
> > > Some interesting usability consideration were raised on the PKIX list
> by Martin
> > Rex in the thread "another attempt to canonicalize local parts" that
> could be
> > addressed by a rechartered WG.  He points out problems with
> > > - Incompletely specified cert chains preventing signature verification
> i.e.
> > missing CA certs and no ability to fetch them.  Fully specifying the
> chains would
> > resolve this.
> > > - Reverifying old "archived" emails on his MUA not possible as the
> certs have
> > expired.  Fixing a verification time on delivery or other scheme is
> desirable.
> > >
> > > -Wei
> > >
> > > On Thu, Mar 10, 2016 at 11:12 AM, Wei Chuang <[email protected]>
> > wrote:
> > >
> > >
> > > On Mon, Mar 7, 2016 at 10:58 PM, Russ Housley <[email protected]>
> > wrote:
> > > I am hearing interest in these topics (a combination of things on this
> list and
> > side conversations).
> > >
> > > (1) Specify the way to use authenticated encryption in S/MIME.  Note
> that it is
> > already done for CMS.
> > >
> > > (2) Specify conventions for AES-CCM, AES-GCM, and ChaCha20 with
> Poly1305
> > authenticated encryption algorithms.
> > >
> > > (3) Specify conventions for using Curve25519 and Curve448 for key
> agreement.
> > >
> > > (4) Specify conventions for using the CFRG chosen curves for elliptic
> curve
> > digital signature.
> > >
> > > (5) Specify a way to use PGP public keys in addition to PKIX
> certificates.
> > >
> > > Anything else?
> > >
> > > While I'm afraid of scope creep as resolving the above would be very
> useful,
> > could also mail header integrity and privacy be considered?  Perhaps the
> WG
> > scope be split into near and long term work to help prioritize.  The
> above five
> > items be categorize as near term and the mail header work considered
> longer
> > term?
> > >
> > > -Wei
> > >
> > > (PS yes there is RFC7508 which is experimental, and does not keep
> private the
> > sender and recipient.  Still it is an improvement...)
> > >
> > >
> > > Is this enough to re-charter the S/MIME WG?
> > >
> > > Russ
> > > _______________________________________________
> > > smime mailing list
> > > [email protected]
> > > https://www.ietf.org/mailman/listinfo/smime
> > >
> > >
> > >
> > > _______________________________________________
> > > smime mailing list
> > > [email protected]
> > > https://www.ietf.org/mailman/listinfo/smime
> >
> > _______________________________________________
> > smime mailing list
> > [email protected]
> > https://www.ietf.org/mailman/listinfo/smime
>
>

_______________________________________________
smime mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/smime