Slamming spamming

I’ve been using POPFile for almost a month now, and am very pleased with it. It’s an open-source message proxy that intercepts all the POP3 e-mail coming into any accounts I have and searches it for spam.

Interestingly, the way it does this is simply by associating content words with either being spam or something else (you can create as many categories as you want, but I just do that sort of binary filtering). Not phrases — just words. (Yeah, I thought it was goofy, too, but, damn, it works.)

Though you can force it to classify certain things as spam (by mailing address or parts thereof, or subject line words), I’ve found I haven’t needed to. Simply by providing feedback over time (“No, this is spam, and this is okay”) I’m presently at >95% accuracy (mostly false negatives about mail being spam) over the course of 1,100 messages during the month (i.e., including that learning period). That’s surprising and gratifyingly high.

Looking at the stats, I can also see that about 16% of the mail I get here at home (vs. work) is spam.

Since POPFile is a mail proxy, it’s usable with any POP3 client. It can be set to either modify x-header information or subject lines, so that your mail client can then push the spam (or other classification) into the folder you like. I still review what gets classified as spam (and vice-versa), at my leisure, but it’s trivial to do.

(I do still use SpamAssassin through my host, but that usually just confirms POPFile’s judgments — it always targets the messages that SA identifies, but usually finds even more.)

As an open source product, it’s free (donations are accepted), and continues to be improved and upgraded. If you’re looking for a relatively simple and quite reliable anti-spam product, and are willing to invest the initial training time, POPFile is definitely worth looking at.