On Fri, 23 Feb 2007 10:49:19 -0500
David B Teague wrote:
> But that doesn't work well when the spam is in an image with
> innocuous text in addition to the image. The innocuous text confuses
> the Baysean data base, making the tool ineffective. Before these
> things began to pester me, Thunderbird's tool worked well, but now,
> it is just a waste. I really want something that will challenge
> everyone not on the whitelist.
I don't know about Thunderbird, but with a Bayesian system
you teach it what is spam and what is not. That means that if you don't
normally receive image-based messages you can teach it to classify them
as spam. I have never had my Bayesian system (Bogofilter) mis-classify
a message because it contained an image. The messages it has
difficulty with are those spam messages that are very similar to the
*normal* non-spam messages you receive. Though after some period of
training it is even very good at dealing with these, in my experience.
YMMV, though, because some people's view of legitimate email can be
quite different from that of others.
ubuntu-users mailing list
Modify settings or unsubscribe at: