I dropped the test setup of Spammassassin on my production server to accommodate a setup of DSPAM. There is no scientific proof that either one is better, at least I couldn’t find it.

One of the major reasons to switch for me was the web interface that comes with DSPAM. With that you can enable users to train their spam filter, check the quarantined messages, identify the false positives (messages tagged spam and that aren’t) and correct them. One other, not so important, reason is that I found. When researching spam filters I read that spammers adapt their strategies to the counter measures spam filters develop. But they adapt only to the most used spam filters. It’s like virus-writers targeting Windows users. At least that is the theory.

I’ve got DSPAM running for two days now, with some hurdles, it’s not running in it’s complete and proper form but it is working. I’ve started with a blank corpus and started training from the beginning and I’m already getting good results. I even opened up my Postfix configuration to be less strict so I receive more spam then normal. What I was unable to achieve with my Spamassassin setup, filter the image spam, is working now with DSPAM. After training DSPAM with three image spam messages it currently blocks them, which is a sign of more promising results in the future.

I will write proper documentation on how to set it up, but before I do that I need to get a proper bullet proof working setup with the web-based administration (which I haven’t looked at yet). If you want to know what I’ve done to get my current setup please read my entries in the forum. Please be careful and only try to do this if you know what you are doing. I don’t understand all the finer details yet, but I’m learning as we go along. I’ll keep you posted on any progress.