Thursday, June 22, 2006

Spam spam spam spam

As well as blocking known spam at the mail server, I run the open source Spam Assassin on my desktop. It is set to throw away everything it is certain is spam, and I never see those emails. However, there are other emails it isn't quite sure are spam, and they get dumped in a mail folder for me to periodically check. Today, I finally got around to doing this for the first time in about six months. After an entertaining half hour skimming the subject lines of almost 12,000 spams emails, I found four false positives.

That's a damn good error rate: less than 0.04%.

However, the false negative rate is much higher, probably around 5% -- despite the spam filtering, I still get about fifty spams a day. I could reduce that by having Spam Assassin be stricter with its filtering, but I'd rather err on letting spams through than tossing away real mail.

No comments: