Re: [rtcweb] Security Architecture: SDES support is a MUST

> -----Original Message-----
> From: rtcweb-bounces@ietf.org [mailto:rtcweb-bounces@ietf.org] On
> Behalf Of Eric Rescorla
> Sent: Friday, July 20, 2012 6:50 AM
> To: Harald Alvestrand
> Cc: rtcweb@ietf.org
> Subject: Re: [rtcweb] Security Architecture: SDES support is a MUST
> 
> The draft currently states:
> 
>      *  The "security characteristics" MUST include some mechanism to
>          allow an out-of-band verification of the peer, such as a
>          certificate fingerprint or an SAS.
> 
> I'd be happy to have us specify a particular fingerprint algorithm
> (presumably
> SHA-256, potentially truncated) but I haven't seen any evidence that
> users savvy enough to inspect the fingerprint at all can't navigate
> more
> than one hash. For example, Chrome's SSL/TLS inspector currently
> displays
> MD5 and SHA-1. So I'm not sure something needs to be standardized here.

Users are unreliable to do this.  

I would look at TLS key pinning or similar techniques.  That is what 
ZRTP is doing to avoid speaking a SAS for every phone call.

Again (for the group), DTLS-SRTP allows us to build identity validation
such as key pinning or a SAS or third-party IdP, but SDESC does not 
allow us to build identity validation.

-d

> -Ekr
> 
> 
> On Fri, Jul 20, 2012 at 6:28 AM, Harald Alvestrand
> <harald@alvestrand.no> wrote:
> > On 07/20/2012 03:18 PM, Fabio Pietrosanti (naif) wrote:
> >>
> >> On 7/20/12 3:06 PM, Harald Alvestrand wrote:
> >>>>
> >>>> Current security definition of WebRTC does not support end-to-end
> >>>> security.
> >>>
> >>> The current security definition of WebRTC (with DTLS) provides
> >>> fingerprints.
> >>> If the application is able to verify those fingerprints, security
> is end
> >>> to end; if it isn't - it isn't.
> >>
> >> The security specification already does specify how the fingerprint
> must
> >> be checked, against a third party system that must be trusted
> (unless
> >> there is some recent update i didn't still checked).
> >>
> >> The way the specification describe fingerprint must be checked, does
> not
> >> enforce end-to-end security but always rely on trusted third party,
> >> being IdP (identity providers).
> >>
> >> The only way to achieve end-to-end security is not to have any kind
> of
> >> trusted third party, as has been already discussed on
> >> http://www.ietf.org/mail-archive/web/rtcweb/current/msg04043.html .
> >
> > That link is your request to have SAS considered mandatory.
> >
> > I believe you got some support for that proposal, or at least some
> kind of
> > availability of digest information that can be verified
> independently, but I
> > do not believe that you got any support for having SAS be the one and
> only
> > definition of "end to end security".
> >
> > Thus, you may get support for your real request, but not for the
> language
> > you use to describe it.
> >
> >>
> >> Until WebRTC security architecture specification does not clearly
> define
> >> a peer-to-peer fingerprint verification system that does not rely on
> >> trusted third party it cannot be considered to provide end-to-end
> >> security.
> >>
> >> Fabio
> >> _______________________________________________
> >> rtcweb mailing list
> >> rtcweb@ietf.org
> >> https://www.ietf.org/mailman/listinfo/rtcweb
> >
> >
> > _______________________________________________
> > rtcweb mailing list
> > rtcweb@ietf.org
> > https://www.ietf.org/mailman/listinfo/rtcweb
> _______________________________________________
> rtcweb mailing list
> rtcweb@ietf.org
> https://www.ietf.org/mailman/listinfo/rtcweb