There have been efforts by legislators to ban spam. These efforts are completely misguided since (for some reason) few lawmakers understand that the internet doesn’t stop at the American border! As a libertarian-minded dude, I also have a problem with banning any speech, even if it is considered spam.

For this reason, I’ve always preferred a technical solution to the problem of spam. After all, geeks invented email. We should be able to clean it up, right?

For a while now I’ve been looking for a good spam filter. Siteseers runs its own mailserver, of course. With that freedom comes the potential for abuse (or another way to look at it is that I own my domain, so I’ve kept the same email address for years). Thus, I’ve been getting enormous amounts of spam.

There are plenty of anti-spam approaches and packages out there. When I ran Qmail, I killed 50% of spam by simply blocking mail connections from Asia. However, I moved to Postfix a week ago, and never got around to figuring out how to use tcpwrappers with it.

That one week of not blocking Asian-originating spam was eye-opening. Fifty to eighty percent of my email has been spam. Since I was considering new ways of battling it, I looked into using Bayesian filtering to filter it for me.

The first Bayesian package I tried was from the open-source luminary Eric Raymond called Bogofilter. After finally putting the parts together to compile it, I fired it up on my mailserver, Maestro.

Boom! It segfaulted. Hey Eric, spend a little less time playing “rock star” and a little more time coding, ‘k?

I looked into other packages, like BMF, but none really did what I wanted. Until I found ASSP.

ASSP is short for Anti-Spam-SMTP-Proxy. It deals with spam in a unique way – stopping it from ever entering my mailbox. It does this by acting as an SMTP proxy between my mailserver and the outside world. When enough of the message has crossed my firewall to make a judgement on being spam or not, ASSP checks it against its Bayesian filters and scores it. If its determined to be spam, ASSP clips the SMTP session right there. The message never gets delivered. Life is good.

If ASSP really can’t tell (and it’s very good. Bayesian filtering is amazingly accurate), it will score it and send it on, letting me decide. I had two spams in my email box this morning rather than two dozen. And I can easily write a rule to check the spam score and deal with the spam accordingly.

ASSP was designed to simply work. Sure, it takes a little prep time to get it going, namely you have to feed it sample emails. Luckily, I’ve been saving my spams since December so I had lots to provide it. Once ASSP learns spam from nonspam, it becomes smart about new messages. The author claims that ASSP could filter spam for a year without updating. Niiiiiice.

All these features, and it also runs under Windows. That’s because it’s written in the ubiquitous perl. So Windows users can get relief, too.

In short, after only twelve hours of running ASSP, I’m in love. I may just keep an eye on its logfile during the day and chuckle at the hapless spammers being stopped in their tracks.