The OPSEC working group convened at IETF 74 on 24 March 2009 at 17:20 PDT.
Note taker: Joe Abley

1. draft-ietf-opsec-routing-protocols-crypto-issues-00

Vishwas Monral presented a summary of the background of this document,
observing that it has been presented in various venues before.

Ron Bonica: five bullet points of the issues you have uncovered?

Vishwas: Things like "all the routers share the same key". What if
  one of them is malicious. In a normal mechanism, the same key is
  not shared amongst all the routers.

Vishwas: Another one is replay attacks. Someone could pump in older
  information and adjacencies could break.

Chris Morrow: most routing protocols I've seen don't require the
  same key everywhere, it can be per-adjacnecy, per neighbour, etc.
  The use of the same key everywhere is an operational convention.
  Also note that the passwords *never* change, since it's practically
  impossible to change them on networks of any size. Having some
  way to change these keys in a reasonable fashion would be terrific.

Vishwas: In ospf we have mechanisms for key rollover.

Chris: Draft, no running code

Vishwas: Correct.

Chris: I do like this, though. The reasons they are the ways things
  are is because of current operational practices as well as the
  tools that are available to the operators.

ron: I also really like this work. a good way to proceed would be to
  list security requirements, e.g. protection from replay, a way to
  change keys, and then do some gap analysis.

Vishwas also presented draft-bhatia-manral-igp-crypto-requirements.

Joel Jaeggli: Note that this is not a wg document at this moment.

Vishwas: Correct. there was some opposition. Not necessarily real
  issues.

Joel: Oh, there were real issues. there is some controversy in
  general within the ietf, cryptographers vs. protocol engineers.

...

Vishwas: Adding a draft for each protocol was inefficient. Decided to
  create a single draft for all IGPs and made a recommendation in just
  one draft. Any changes could be made in one place rather than being
  made in many documents. Did something similar in IPSec.

Ron Bonica: MAC algorithm that is secure today won't be secure
  tomorrow.  What concerns me is that we've been dealing with
  authenticaiton algorithms that support 96 bit macs. If we need a
  longer MAC in the future, do we need to support it in the routing
  protocol?

Vishwas: The way is-is was, it was written with the thought that
  there would be only one, MD5 hash authentication.

Brian Weiss: Usually we have a 96 bit mask because we truncated the
  actual hash to 96 bits.

Joel: This is important work, in the interests of not being caught
  by surprise in the manner that (e.g.) some operators were when
  they had to suddenly do widespread deployment of TCP/MD5 with
  BGP.

Pat Kane: the question before the security people is how do we move
  to SHA-3. One thing that came up today at liunchtime is that it's
  a problem to make a statement that MD-2 or MD-5 is not good for
  anuything. Hashes are used for different things, not just crypto.

Joel: ...

Pat: Yes, but also worried about the other end; ssh doesn't support
  md-7 yet, omg, etc.

Richard Grey: there's an effort between routing and security areas
  under name of kmart which addresses most of these problem areas
  one way or another. There's a mailing list, kmart@ietf.org where
  there has been discussion. Not a wg, but individual submissions
  in a mailing list.

Joel: Ops is not the place to do the protocol work.


2. Practical Active Attacks on IGPs that Some Folks Might Not Realise
Exist (RJ Atkinson)

Ran was not available.


3. draft-ietf-opsec-blackhole-urpf-02

Deferred.


4. draft-ietf-opsec-ip-security, draft-gont-tcp-security,
  draft-ietf-opsec-icmp-filtering, Presented by Joel.

Joel: Thought that whether tcpm might be another place to socialise
  draft-gont-tcp-security

Ron: at least, tcpm. 


5. Document Status Summary

Joel: Document status. Vishwas/Manas have one document that is not
  on the list. What is here are warren kumari's blackhole URPF draft,
  revised in the last month, believed to be very close to WGLC as BGP.
  Major change was to remove the codepoint request for a special
  community for a blackhole.

Joel: ICMP filtering document shold be revised shortly, recently
  expired. Beyond that document state hasn't changed that much since
  last meeting. Hope we do WGLC on two documents shortly, one on
  opsec routing protocols crypto issues, and either accept or
  encourage the other document to be accepted as a wg draft, or
  recast as an individual submission.

-end-