A couple of months ago, I switched to using Outlook 2003 on my Windows XP system at work. Outlook 2003 offers quite a few nice feature enhancements over its predecessors, but one of them that interested me is the junk mail filtering. I configured the filter and found that it worked okay, but I was still getting quite a bit of junk email. Then, a couple of weeks ago, a new class of junk email started hitting my Inbox. These new emails replace common characters of the alphabet with look-alike symbols and accented characters. The result was that I started receiving a significantly larger number of spam emails.

I decided that I had had enough and I went looking for a Bayesian junk email filter. I talked to a couple of colleagues and settled on SpamBayes. I usually delete spam emails from my junk email folder, so, unfortunately, I didn't have a lot of emails to train the system with. But, even with a small set to train from, SpamBayes took a significant bite out of the junk emails that made it to my Inbox. Over the past couple of weeks, since I installed it, it has continued to improve and the number of junk emails that make it through the net has dwindled to a trickle.

SpamBayes represents a significant improvement over Outlook's built in junk email filter, but it still isn't perfect. The character substitution technique being used by spammers now is a pretty effective way to bypass the Bayesian technique and I'm sure SpamBayes will be upgraded soon to take this technique into account. But, be that as it may, it is still doing a great job. 10:32:40 AM    comment []