Bayes Poisoning? Spam with negative BAYES Scores - ahhhh

Mark Nienberg mark at TIPPINGMAR.COM
Wed Dec 24 22:52:42 GMT 2003

On 24 Dec 2003 at 10:36, Nathan Johanson wrote:
> I do like the idea behind reducing the Bayes probabilities. Please let
> me know how this works for you. I'm curious if it's enough to fix the
> problem, or if it impacts your filtering in some other unforseen way.

I've had this in place for about a week now, and I haven't had any problems.  The
only potential problem I can see is if you regularly receive legitimate messages that
for some reason score higher than your spam threshold and require the Bayes
negative score to reduce the total below the threshold.  It's a trade-off between
tagging more spam and the possibility of some false positives.  I haven't seen any
false positives yet.

That said, the SpamAssassin purists might be horrified at this kind of tweaking,
because it messes up the careful statistical analysis on which the scoring is based.  I
understand that, and generally agree with it, but I guess there is nothing wrong with
trying to tweak SpamAssassin to work best with the type of mail that my site

I can't say that this has fixed the problem entirely though, because this type of spam
often doesn't trigger enough other points to go above the threshold, even without the
Bayes score to bring it down.


More information about the MailScanner mailing list