Clearswift’s Interesting Move Into Auto-Learning Spam Filters

We recently wrote that statistical filters such as POPfile are extremely accurate when used correctly, but too hard for most people to use. That’s why we’re excited by the potential of Clearswift‘s new product line. It takes the underlying technology from POPfile — in the form of the commercial polymail library — and adds the capability for automatic learning. Clearswift uses the library to implement statistical spam control without the need for end users to train the filter. Here, the filter is trained not by users flagging errors, but by calculating how confident the filter is that a message is spam or not.

It decides its confidence by using a cocktail of techniques, such as IP blacklists, call-to-action blacklists, and automatic whitelisting. If this cocktail indicates that a message is probably spam and the statistics indicate that as well, then the filter will train itself that messages that look similar are also likely to be spam. Similarly, it will auto-learn what legitimate messages look like.

Richi Jennings

Post a comment

You must be logged in to post a comment. To comment, first join our community.