Hello,
> Perhaps you should look into bayesian spam filtering, like ASSP. I found
> it to be very effective.
I found it to be very ineffective especially with image spam and also to
cause a lot of false positives.
> A spamfilter will always be high maintenance,
> if you want it to be effective (for example you want to keep teaching a
> bayesian filter). I believe the quality of a spamfilter depends greatly
> on the quality of the admin working with it.
I don't want to pay a sysadmin a train a spam filter. I want an antispam
solution, that is good out-of-the-box and is trained by someone else
(e.g. honeypots of the vendor). We are managing about 40.000 email
addresses and currently > 1.000.000 messages daily (whereas >93% is
spam). Thats nothing, that can be trained manually any more.
