� Getting ready for postseason | Main | Countdown to 2nd Dan �

October 1, 2005

Cure for spam

I've been using Spamassassin for quite awhile on my server, but never really understood its true power until I began playing with the Bayesian classifier. SA uses the 'sa-learn' utility which deploys a Bayesian algorithm. You need not be a mathematician harness the power of the Bayesian classifier, but for those interested in an explanation of the mathematics I've included it here.

Essentially, you simply have to make sure that you're running SA on your mailserver. If you don't have access to your server, ask your ISP to install it. IMHO, it is far better than the crap that some of the commercial packages that are in use today.

On my server, I simply run the 'sa-learn' script in this fashion;

sa-learn --showdots --mbox --spam ~/mail/bkaeg.org/xxx/spam (Learns about the bad stuff)
sa-learn --showdots --mbox --ham ~/mail/bkaeg.org/xxx/inbox (Learns about the 'real' mail)

I get virtually zero junk, as I run a cron routinely to improve the learning. Basically, the algorithm learns about the behavior of good email versus the hated spam.

Posted by AG at October 1, 2005 11:45 AM