Re: [OSPF] Supporting Authentication Trailer for OSPFv3

[Curtis Villamizar <[e-mail]>]

In message <[e-mail]>
Acee Lindem writes:
>  
> Hi Srini,
>  
> The fact that graceful restart will be more difficult is part of the
> cost of implementing this draft. One of the jobs of the OSPF WG is
> determining whether the "medicine is worse than the disease". In this
> case, the disease is well-timed replay attacks and the medicine is the
> proposed solution.
>  
> Thanks,
> Acee


Acee, et al.

I hope no one minds that I trimmed the rest of the context.

It should be possible to on startup negociate a new initial sequence
number through an exchange that involves the exchange of an encrypted
or authenticated challenge using a shared key.  This would involve
additional protocol exchange which is in neither of the security
drafts being considered, but either could be changed.

Do so would allow the replay attack problem to be addressed without
creating a new problem due to forgetting the last sequence number that
was used after a gracefull restart wakeup.

If we are going to go to this extent, adding a negociation step, then
you might also want to add an option to exchange an encrypted session
key to avoid an attack where enough "in the clear" information is
authenticated to guess the key in use.

I can barely keep up with IETF email so I'd rather someone else pick
this idea up if its thought to be a good idea (maybe it isn't).  Now
back to lurking.

Curtis