Fuzzier Spam

SpamButcher uses a fuzzy logic engine to distinguish spam from non-spam email messages. When it was first released, it was the first commercial spam killing application to use fuzzy logic. In subsequent years, any number of spam blockers have been released using similar if not near-identical techniques.

The core of fuzzy logic is the fuzzy set. In contrast to non-fuzzy sets, an entity may be a member of a fuzzy set, and non-member of the same set to some degree. In the case of SpamButcher, there is usually only one set; the spam-mail set.

Every message is a member of the spam-mail set to a certain degree. If a message's membership reaches a certain percentage, it's considered to be spam, and is filtered out to the intercepted messages folder. In effect, set membership is an analogue data point, but filtering is objective. A message is either blocked, or it isn't.

In recent months, severely aggressive spam messaging campaigns have begun to comprise a larger portion of all junk email. Analysis has shown that these messages often contain multiple identifiers that can be used to recognize them as part of a given campaign. These identifiers are often however, not suggestive of spam in general.

As a result, the anti-spam email software now attempts to determine if a given message is a member of certain specific spam-campaign sets. If a message meets enough criteria, it will be blocked. If a message is a below the threshold, it won't be filtered, and those criteria found won't effect the message's membership to the general spam-mail set.

Back