-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Jeffrey 
Haas
Sent: Friday, April 01, 2011 9:12 AM
To: [email protected]
Subject: [sidr] IETF 80 - suggestions related to expiry time and BGP 
implementation

2. Short expiry times are an attack on the routing system, especially boxes 
with slow signature processors.  Routes that will expire
"soon" should be refreshed with enough time so that receiving peers can take 
their own sweet time to validate that a new valid path
has been received in spare cycles.

[WEG] Along those same lines - I think we need a global minimum value for 
beacons to reduce the exposure to this DoS attack (on the
validating router) vector. We already have scaling issues when too many  eBGP 
peers set their BGP timers too aggressively on the
same box, and that is at least locally controllable by enforcing a minimum in 
the negotiated session. There's no negotiation
happening here, and therefore no way to protect those routers, short of 
ignoring updates, which generates stale routes and probably
churn (depending on the policy).
In fact, this starts looking a lot like the Route Flap Dampening problem we're 
trying to solve elsewhere in IETF - a very small
amount of the prefixes might be responsible for an overwhelming majority of the 
updates to the detriment of the routing system's
performance. Even if it's just a refresh, there's a non-trivial impact if too 
many people start being overly paranoid about their
expiry times. 

But by the same token, being too conservative with the expiry means that there 
is a longer period of exposure where a replay would
be effective, and the value of implementing the overhead at all starts becoming 
questionable. How long is too long for a replay
attack to go unnoticed? I'd bet that a lot of the folks worried about this 
would answer in minutes, while those concerned primarily
with the hardware in their routers would answer in hours...

This is an area where any info we have about the prevalence of replay attacks 
and their characteristics would be really helpful in
determining how much risk we are preventing by addressing this specific attack 
vector vs how much overhead we're generating with
this expiry machinery.

Thanks,
Wes George

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr

Reply via email to