Greylisting, and making use of the Bayesian filters in SpamAssassin, is what we’re using on our servers…it isn’t perfect, but, it’s not awful either.
Spam is a real problem that gets worse every year, unfortunately. It also gets more sophisticated every year. I’ve noticed my gmail account gets almost no spam, but has a lot of false positives, so the tradeoff is hard even for an entity with the resources of Google.
It might be worth looking into how to make the Bayesian filtering work better in Virtualmin systems (e.g. by making it easier for all users to contribute to the spam database).
Also, DKIM and SPF are becoming widespread enough to be useful filtering features. I haven’t done a lot of experimentation with this, but making them part of the weights that SpamAssassin uses is probably a good choice. I’ll ask Eric and Jamie if they’ve got any experience with this. I’d be curious if other users have worked with it, as well.
thanks … this is a little overwhelming for me! can i install DCC, Razor and Pyrzor through the software packages section of virtualmin? will configuration all take place at the command line or can any of these packages be configured through the spamassasin/postfix sections of virtualmin?
Re: can i install DCC, Razor and Pyrzor through the software packages section of virtualmin. No I don’t think so which is why I did it this way.
It is only the installation you use the command line the rest is editing the /etc/mail/spamassassin/local.cf and /etc/mail/spamassassin/.razor/razor-agent.conf files and then restarting spamassassin as well as opening the ports in the CSF firewall if you use it which you should be doing.
Try it on a test machine first so that you are more comfortable about doing it.