Re: [abfab] Why we need Nico's text

[Klaas Wierenga <[e-mail]>]

On 8/25/10 3:29 PM, Sam Hartman wrote:

Sam,

Thanks for the explanation, very helpful to me at least!

Klaas

> I'm noticing that the folks who have worked with GSS-API seem more
> comfortable with Nico's text than the rest of the community.
> Let me go through Nico's requirements and explain why they're needed in
> order to accomplish our goals.
>
> * Mutual authentication: Required by RFC 4462 in order to work with SSH
>    and RFC 5801 for XMPP, SMTP, IMAP, LDAD, and the like.  You introduce
>    MITM attacks for CALDAV, WEBDAV, ATOM and the like if you use GSS to
>    authenticate without mutual authentication.
>
> * Channel binding: required by RFC 5801 if we want to support IMAP,
>    XMPP, LDAP, and the like.
>
> * Per-message security services (gss_wrap, gss_get_mic, the like):
>    Required by RFC 4462 if we want to support SSH. Required by
>    NFS. Required by CIFS.
> \
> * Host based service naming: Required by RFc 4462 for ssh, RFC 5801 for
>    XMPP, IMAP, SMTP, and the like. Required for CALDAV, WEBDAV, ATOM, and
>    the HTTP protocols. Required by NFS. Probably required by CIFS
>    although that's more complicated.
>
> * gss_pseudo_random: Required by some of the JANET applications and by
>    RXGK, the AFS distributed filesystem security class. This is less of a
>    requirement than the previous items, although once you have
>    per-message security, gss_pseudo_random is trivial.
>
> * key confirmation: Required for channel binding to provide meaningful
>    security.
>
> So, if we don't meet Nico's requirements then we cannot actually work
> with most of the applications we want to work with.
>
> However, for me, that's not the most important motivation.  I've been
> working on security systems that are more resistant to phishing than the
> web for a long time. These technical requirements will meet the higher
> level requirements necessary to provide the technical aspects of good
> phishing defense. It happens these requirements are a fairly good
> mapping of the most important issues discussed in
> draft-hartman-webauth-phishing to non-web applications.  So, that's why
> I personally care a lot about these issues and whether these are going
> to be part of the solution significantly affects how good of an idea I
> think doing the work is.  I can totally appreciate that someone might
> disagree with me, but I hope you can understand why I care about the
> issue.
> _______________________________________________
> abfab mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/abfab