The Secrets of Spam Filtering

Some users have understandably expressed frustration regarding how SpamButcher's email filtering is effectively a black box. While it's possible to create your own good and bad phrases, its core set of filters is hidden and not adjustable. There are a few reasons for this.

First, SpamButcher uses a technology called fuzzy logic to identify email messages that are most likely unwanted. The fuzzy logic engine does ultimately rely on text phrases. However, those phrases have additional information attached to them to help indicate their relevance. It would be possible to create a user interface to the database, but most users would likely be confused by the spam mail blocker system.

Second, spammers love it when spam filter authors publish their findings. There are many open-source related databases of Bayesian spam filtering statistics that are actively in use. If a spammer wants to, they can easily check to make sure their message doesn't contain any offending phrases. In fact, I suspect there are several spam sending packages that pass messages through such spam programs to verify they won't get caught.

Spammers can of course also use SpamButcher to verify they're messages will sneak past. The difference is, they won't be able to tell why SpamButcher caught their message if it does. This makes it much more difficult to evade detection.

Back