Months ago, I threw in the towel with my old Spam filter, SpamFire. It was apparently written in RealBasic, and so it was slow and constantly got “stuck.” I’d rather have spam than have to be constantly nursing a memory hog, finicky application.

Finally I got fed up with my spam problem this morning and I’m trying something new. 95% Spam is just too much to handle.

Spamato

Spamato is an interesting Spam filter solution. It works by having a set of filters which all look at the mail and then sort of vote on whether it is to be considered spam. When training the filter, I notice that commonly three of the eight filters will hit most of my obvious spam. After training is done, presumably more will hit.

I’m using the Spamatoxy mail proxy filter, rather than the Thunderbird plugin because of OSX compatibility problems. Still, it is quite easy to set up, and training is only one step more difficult than it would be with a full integration.

Training

I have seven accounts I actively check. One of them is IMAP, the rest are POP. So, I set up my IMAP filter to use a spam folder as a “smart mailbox.” All IMAP spam goes into that folder, and if I manually drag a file there, it gets reported to Spamato as a spam, the reverse is true as well, if I drag one out, then it is “revoked” and is not considered spam.

To train it, I just dragged a few hundred junk mails from my local junk mail directory into my IMAP inbox. I made sure all were marked *spam*, then I cleared my spam folder. to train for good mail, I dragged a few hundred good mails from local folders to the IMAP inbox, and then made sure none of them were tagged as spam. A few dozen were, but it was no bother to revoke the spam dermination.

After doing that, my Bayesian filter should be active. I’m curious to see how much better the system gets with the Bayesian filter active.

[tags]spam,spamato,spamatoxy,spamfire[/tags]