On Friday, August 16, 2002, at 01:10 PM, Chuq Von Rospach wrote:
>
> <http://www.paulgraham.com/spam.html>
>
> It's a new technique for identifying spam. The more I look into the
> details,
> the more I think we have the "anti-spam killer app", becaues it tunes
> itself
> to the individual (or site), adapts as the anti-spammers adapt, and the
> technique used is fairly easy to implement and damn difficult for a
> spammer
> to avoid....
People have been talking about Bayesian filtering for spam for quite a
while. It is definitely very promising. But the easiest way to defeat
it -- which also works against sophisticated pattern-matching algorithms
as in SpamAssassin -- is simply to make the payload of your spam an
image, which cannot be turned into lexical tokens for Bayesian
analysis. Even if a clickable link is appended, that will not provide
sufficient information to declare something spam, since people send URLs
to each other all the time. You can simply assert that a message
consisting only of an image and a link is spam, but that's not Bayesian,
that's just a semantic pattern.
--
Michael C. Berch
mcb@postodern.com
Follow-Ups:
References:
|
|
