[118256] in Cypherpunks

home help back first fref pref prev next nref lref last post

Re: KISA Attack

daemon@ATHENA.MIT.EDU (Greg Broiles)
Thu Sep 23 00:38:44 1999

Message-Id: <4.2.0.58.19990922205727.00bc88c0@mail.wenet.net>
Date: Wed, 22 Sep 1999 21:05:44 -0700
To: Sean Roach <roach_s@mail.intplsrv.net>, cypherpunks@algebra.com
From: Greg Broiles <gbroiles@netbox.com>
In-Reply-To: <3.0.6.32.19990923074752.008226b0@mail.intplsrv.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Reply-To: Greg Broiles <gbroiles@netbox.com>

At 05:47 AM 9/23/99 , Sean Roach wrote:

>Excuse me.  This will undoubtedly show my ignorance, but.
>What would that serve?  If they are tying up your lines to prevent
>your serving others, what advantage would holding the line open for
>them serve?  Unless the bottleneck is processor time or file access
>time and not the network connection.

The underlying assumption is that the bottleneck is a single process on the 
other machine which is tasked with gathering web pages (maliciously or not) 
- that seems like a reasonable assumption, because a single-threaded 
program which visits pages serially is easier to write (and less likely to 
consume scarce local resources) than a multiple-threaded spider or attack 
tool. The latter is, of course, easily possible, but requires more effort.

I suggested the above because the behavior John described sounded like a 
normal spider, but operating without a rate limiter - the existence of 
malice on the part of the operator being unimportant from a technical 
perspective. If that single process can be held open but inactive the first 
time it makes a connection, and can be slowed down so that it completes one 
request per minute (or hour, or day) rather than one per second, the 
bandwidth (and other resources) used by the spider will be much less.

The important distinction between the "teergrube" approach and the "small 
result" approach you suggested is that the teergrube completes very slowly, 
so the cycle time for the remote machine is very slow. Small but quickly 
completed transactions allow the other machine to immediately recycle, to 
reattack your machine or someone else's. The "teergrube" tactic might make 
more sense if it was called a tar baby instead.


--
Greg Broiles
gbroiles@netbox.com
PGP: 0x26E4488C


home help back first fref pref prev next nref lref last post