Re: We have a firewall (was Re: Pakistan government orders ISP service level agreement)

["Christopher L. Morrow" <chris () UU NET>]

On Tue, 6 May 2003, Phil Rosenthal wrote:

> On 5/6/03 7:51 PM, "E.B. Dreger" <eddy+public+spam () noc everquick net> wrote:
>
> > SD> Date: Tue, 6 May 2003 19:28:48 -0400 (EDT)
> > SD> From: Sean Donelan
> >
> >
> > SD> The Pakistan Telecommunications Company Ltd has aquired a
> > SD> firewall to solve the DDOS situation impacting Internet
> > SD> service in the country.  An unnamed security advisor asserted
> > SD> the proper use of a firewall would control the DDOS attacks
> > SD> and prevent hacking.
> >
> > Now the DDoS melts the pipes _and_ the firewall.  I'd like to
> > know if said "consultant" ever considered recommending the PTC
> > contact their upstreams for help with backtrace/blocking.  Anyone
> > with a modicum of clue (or Google access) should figure out that
> > one...
> Not every upstream is as clueful as Uunet, and not every noc employee is as
> clueful as Chris and Brian at UUnet.

oh oh... there are quite a few folks who work, including the NOC here, to
make Brian look good :) (and me, but mostly brian looks good)

> It has been my experience that most upstreams have no concept that they CAN
> backtrace, and generally have no interest in helping you do it.  I'm not
> mudslinging here, so I won't say who my experience is with, but a few
> transitless/near transitless upstreams I've dealt with were most unhelpful,
> either because they didn't know how to help, or worse, they did know how to
> help and didn't care.

Unfortunately this is the case at times, I will name some good names
though, C&W/Qwest/Verio/Sprint/ATT among the larger carriers I've recently
dealt with, in the US. Francetelecom and SwissComm atleast external to the
US are also quite helpful these days. It seems that atleast all of these
folks have been active in stopping many recent attacks. There are some
others that don't seem quite as helpful, but that number is getting
smaller.

> And, depending on the nature of the DDoS attack, perhaps it isn't related to
> saturation, but rather to overloading router processors, or something else
> that can effectively be filtered customer-side?

There is a fine balance that has to be struck... killing a provider side
router and N customers or degraded service for a single customer who can
still filter their side of the link :( sometimes people aren't happy with
the response.

> Our policy as of late has just been to make sure we have equipment on our
> side fast enough to filter at wire speed, and get enough capacity to our
> upstreams that it is signifigantly unlikely that anyone could generate
> enough traffic to saturate it (in which case, we would have no choice but to
> ask carriers to filter, and backtrace).