The Sonoracomm howto is more up-to-date:
http://www.sonoracomm.com/support/19-inet-support/49-spam-filter-configuration-for-sme-7I have 3 SME servers setup as spam filters.
I do the following on each server:
Enable
DNSBL and RHSBL filtering
Enable
Bayesian AutolearningSetup 'LearnAsSpam' and 'LearnAsHam' using either these manual instructions:
http://bugs.contribs.org/show_bug.cgi?id=1701#c36 or using the
Learn contrib (alright - I always use the manual instructions, and have never personally used or tested the Learn contrib -- but it should work fine!)
Since my users can't easily move emails into the 'LearnAsHam' or 'LearnAsSpam' folders, I create public folders on the Exchange Server for 'LearnAsHam' and 'LearnAsSpam', and teach my users to drag or copy appropriate messages into each folder.
I then create a mailbox on the server called 'spamfilter' and do one of the following:
- Add an IMAP account into Outlook for a few select users, and allow them to train the spam filter by moving the user-filed messages from 'LearnAsHam' and 'LearnAsSpam' in Public Folders into the corresponding folders for the user 'spamfilter' on the SME server.
- Install and configure
ImapSync to automatically suck the messages from the exchange server into the LearnAs... folders on the SME server every 15 minutes or so.
I monitor the SME spam filtering behavior using
qplogtail and
qploggrep.
I use rather permissive settings for DNSBL and RHSBL servers to avoid blocking too much traffic (The default settings are very likely to generate false positives, in my experience). My SBList is set to
rhsbl.sorbs.net, and my RBLList is set to
zen.spamhaus.org:combined.njabl.org.
Following these procedures, I have never needed to whitelist anything on the SME server itself. When users start baying about "I need to whitelist xyz!" I teach them how to use 'LearnAsHam'.