Spambayes Anti-Spam as POP3 Proxy

Original date: May 27, 2006 5:01 AM by sonnik

I'd have to say that E-mail spam isn't as big of problem as it was in 2004 or so. Many ISPs have installed software to do the work for you, or some desktop solution has probably been installed on your new computer to assist. However, E-mail spam can still be a problem - and while clever solutions have came and went, the principles of 18th century mathemetician Thomas Bayes seems to still work best.

So, we have SpamBayes - an open source solution to analyze and label your incoming mail. This Open-Source package includes a few solutions, such as an Outlook Plugin or a small Proxy server.

For me, I went with the Python-based sb_server (sb_server.py). Documentation is more than adequate at the SpamBayes web site, if you'd like to give this a shot.

As I've mentioned before, I have a small always-on Linux machine performing tasks for me. Anti-spam filtering is one of those tasks. The sb_server.py actually provides a handy configuration interface via a web browser, to deal with the nitty-gritty details of training the script just what exactly you consider Spam. (Note that for security reasons, the default installation only allows localhost for the web interface. You can adjust it to your home network's subnet if you desire.)

Once you've configured and trained sb_server.py a bit, it will start acting as a proxy between your ISP's POP3 E-Mail servers. The messages will then have a few additional headers scoring the E-mail as "HAM" or "SPAM" which you then filter in your typical E-mail client. I use Outlook's filters to bin the spam into a folder automatically. I do a quick browse to make sure I'm not deleting anything important.

You'll note that SpamBayes likes to be properly trained, and offers you a few mechanisms for making sure that 'training' continues during it's life. This is handy, because spammers are always changing their methods. Fortunately, SpamBayes seems to remain up to the task.

When starting out, it's best to have a couple hundred messages of each variety. If you're using Outlook, export your messages to a .PST file, which you can then convert to the MBOX format that sb_server.py understands. ol2mbox will perform this task for you.