Re: [trill] Alvaro Retana's Discuss on draft-ietf-trill-directory-assisted-encap-10: (with DISCUSS)

[Donald Eastlake <[e-mail]>]

Hi Alvaro,

On Mon, Mar 5, 2018 at 2:32 PM, Alvaro Retana <[e-mail]> wrote:
> Alvaro Retana has entered the following ballot position for
> draft-ietf-trill-directory-assisted-encap-10: Discuss
>
> ...
>
> ----------------------------------------------------------------------
> DISCUSS:
> ----------------------------------------------------------------------
>
> I have significant concerns about this document; as currently written, I
> believe the technology is underspecified and can cause significant damage to a
> DC network where it might be deployed.  I am then balloting a DISCUSS.
>
> The document (including the security considerations) is written assuming that
> the TRILL-ENs can be trusted (and are not compromised), and that the directory
> information is accurate.  However, I believe there are several cases that have
> been overlooked.
>
> (1) There aren't any basic safeguards specified to at least make sure that a
> TRILL-EN is doing the right thing (or something sensible).  For example, what
> if the Ingress RBridge Nickname field in the TRILL header doesn't correspond to
> the first rBridge at the domain boundary?  Should that frame be accepted?

This concern doesn't seem very different from the general insecurity of Layer 2.

If a link has no TRILL router adjacencies, then I guess it it
reasonable to check that the ingress nickname is that of the receiving
RBridge but it gets harder if you have a mixed link with both end
stations and TRILL routers (probably rare in deployments but, on the
other hand, in TRILL a "link" can be a bridged LAN...).

> (2) rfc8171 talks about issues with incorrect directory mappings.  Consider the
> case where a TRILL-EN uses (on purpose!) an incorrect mapping.  That "can
> result in data being delivered to the wrong end stations, or set of end
> stations in the case of multi-destination packets, violating security policy."
> [rfc8171]  How can this risk be mitigated?

I'm not sure how using "an incorrect mapping" is that different from
just using whatever nicknames it feels like...

> I don't think that there are easy mitigations for these issues, but at least
> mentioning them so that operators are aware of the risk would be enough to
> clear this DISCUSS.

It is certainly reasonable to mention these. One possible mitigation
is the use of end-to-end encryption.

Thanks,
Donald
===============================
 Donald E. Eastlake 3rd   [phone] (cell)
 155 Beaver Street, Milford, MA 01757 USA
 [e-mail]