source filtering

Wed Jan 13 09:36:58 UTC 1999

Hi.

Yep, you are right, you can protect yourself from smurf with a little 
help of your upstream provider (CAR on upstream link and autho-installed 
filter working by netflow analyser). What you can't do and what's much 
more dangerous is netfloods by IP fragments with random -generated SRC 
addresses and a lot of small packets - they flood WFQ queueries, overrun 
netflow accounting, waist CPU poewr of the routers and (amazing) they 
often can't be filter out at all (it's true for the CISCO in some cases).

Again, the only _SOLID_ solution for the future internet is strict SRC 
filtering over all ISP. I note - ALL, may be it should be some kind of 
_open policy_, may be something else.

Internet have not other choise to survive.

On Tue, 12 Jan 1999 prue at ISI.EDU wrote:

> Date: Tue, 12 Jan 1999 15:54:37 -0800
> From: prue at ISI.EDU
> To: nanog at merit.edu
> Cc: Prue at ISI.EDU
> Subject: Re: source filtering
> 
> On Tue, Jan 12, 1999 at 01:11:09PM -0500, Steve Gibbard wrote:
> ==>On Tue, 12 Jan 1999 danderson at lycos.com wrote:
> ==>
> ==>> I'm not sure what the big issue here is with the smurf attacks. If you set
> ==>> up some kind of access list that disables incoming icmp traffic, then turn
> ==>
> ==>That breaks path MTU discovery (see RFC 1435 for more info on that), among
> ==>other things.
> 
> 
> What I have done is write a C program which looks at netflow records
> as they come in,(close to real time).  It checks the source port on the
> Cisco, the flow record identifies, and compares the source address to a
> list of acceptable source addresses for validity.  I give the program a
> list of permit and deny like records with IP address and mask along with the
> ordinal port number.  Any that are not legit get printed out.  You can
> then beat on your customer until they clean up their act.  If they
> don't stop right away you can always put in a filter on the interface
> til they do.  We have pretty strong contract verbage which permits us
> to cut people off if their connection is being used for predatory
> activity.  We haven't had to resort to that yet but being able to say
> "Fix it or I turn your connection down til you do" is very effective.
> 
> The nice thing is that it doesn't slow the router down at all, as long
> as you are doing netflow anyhow.   I modified a copy of fdget which
> is Ciscos netflow demonstration software they have available via ftp.
> That made the programming task easy for a non programming type like
> me.
> 
> Walt Prue
> 
> 

Aleksei Roudnev, Network Operations Center, Relcom, Moscow
(+7 095) 194-19-95 (Network Operations Center Hot Line),(+7 095) 239-10-10, N 13729 (pager)
(+7 095) 196-72-12 (Support), (+7 095) 194-33-28 (Fax)