Problem solve Get help with specific problems with your technologies, process and projects.

Bayesian filters

A look at how Bayesian filters work and where you can find them.

Spam is a problem that isn't going to go away anytime soon. Even if governments legislate against it, Spammers ability to adapt their messages and to get unsuspecting users with broadband connections to broadcast spam (Zombies) makes it a very difficult proposition to prevent unwanted messages completely. Chances are you have already filtered your e-mail, and that your ISP is also actively filtering your e-mail even before it gets to you. The problem is that Spammers are adapting to filters, and getting around these types of systems.

Over the past year companies have begun to introduce anti-spam products based on Bayesian filters. A Bayesian filter evaluates the content of a message and scores it based on an algorithm, typically from 0 (not spam) to 100. You set a threshold of what score you want to filter against, and anything higher than that score does not come into your Inbox. That threshold is usually set at a default of 50.

Recently Eudora 6.0's paid mode began offering SpamWatch, which incorporates a Bayesian filter. If, for example, you move all of your filtered spam into the new Junk folder, you already have a good start on getting the filter to be effective. A Bayesian filter is adaptive. So as you continue to designate items as junk, moving them to the Junk folder and setting their scores to 100 or from the Junk box, unjunking them, setting their scores to 0 and returning them to your Inbox, the Bayesian filter gets better. This newsletter you get might be junk to one user and a treasure to another, but now the end user can define it. While there are methods for defeating Bayesian filters, they aren't foolproof, and messages of those types can often be prefiltered by other systems in use.

You'll find a description of how Bayesian filters work at: and at Products with Bayesian filters are found for all e-mail clients, and include: SpamBayes, Spam Bully, Spam Combat, Mozilla's spam program, ResponZe, and others. A compiled list may be found at:

Barrie Sosinsky is president of consulting company Sosinsky and Associates (Medfield MA). He has written extensively on a variety of computer topics. His company specializes in custom software (database and Web related), training and technical documentation.

This was last published in December 2003

Dig Deeper on Network management software and network analytics

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.