[lug] Stopping the New Generation of Spam

[Sean Reifschneider]

On Tue, Dec 05, 2006 at 06:32:03PM -0700, Daniel Webb wrote:
>Now that I think about it, I'll bet you're right: a Markov classifier would
>have no problem detecting that the message was *too* random.  I'm surprised
>they haven't just started lifting paragraphs from Wikipedia or random web

They have.  The ones I'm getting the most of obviously are pulled from real
web sites.  A sentence reads correctly, and then the next sentence is also
correct, but drastically different content from the previous.  Something
along those lines.  So they're taking parts from different pages and
plugging them together.  For example:

   But I don't write for you.
   and maybe life in general. The Viiv architecture and its attached software
   should handle all the details of ensuring I can play back my copyrighted
   standard of choice in my manner of choice. You have to stroke its hairy
   insides to get it to luminate. That's wide-angle distortion.
   No jaded cynicism necessary.
   Arguably the thinnest player from Samsung, if not anyone, that we've seen.
   You are obviously not a procrastinator and thereby a potential robot.

Sean
-- 
 Follow your bliss.
                -- Joseph Campbell
Sean Reifschneider, Member of Technical Staff <jafo at tummy.com>
tummy.com, ltd. - Linux Consulting since 1995: Ask me about High Availability