Bayes Poisoning - How to combat ?

[Dhawal Doshy]

    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "US-ASCII" character set.  ]
    [ Some characters may be displayed incorrectly. ]

Kai Schaetzl wrote:
> Dhawal Doshy wrote on         Thu, 17 Mar 2005 09:58:58 +0530:
>
>
>>since bayes
>>is no more as effective as it used to be.
>>
>
>
> Oh, it is. Nearly all of our spam gets "bayes_99ed". It's the single most
> effective rule. It depends on your bayes db. F.i. ours has tokens back
> more than 12 months. If you follow the sa-talk list you will also see that
> there are high doubts that "bayes poison" actually works. If your bayes db
> fails on many spams than this can have many causes f.i. you didn't learn
> enough to it recently, or the wrong stuff or you are losing tokens too
> quickly by expiry or your customers do get a lot of mail which contains
> tokens which are also in spam or ... "bayes poison" is just the least
> worse explanation.
>
> Kai
>

No doubt bayes still works effectively (comes 2nd after SURBLs for me)..
just that SURBL is the *current killer tool* for fighting spam with very
(very) low FPs. And bayes can still be fooled once in a while OR needs
to learn sometimes.

As for 'bayes poison' i have never seen it happen but I rather use the
sare poison ruleset than take a chance.

btw no mails on the list for more than 24 hours.. is the SPAM PROBLEM
finally over?

- dhawal

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ (http://www.mailscanner.biz/maq/) and
the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).

Support MailScanner development - buy the book off the website!