SpamButcher - Behind the Cleaver

The antispam filter software would isolate a handful of most likely unique strings and query various spam reporting groups to see if they had already been sent to users.

Some initial manual testing of this technique looked promising. It seemed quite possible for a spam eliminator to identify a large portion of spam this way.

Unfortunately further research suggested there would be unacceptable performance implications - so the idea was set aside.

Several years back I'd read the book Fuzzy Thinking by Bart Kosko. It described a system of reasoning based on "fuzzy sets" - everything could be described as belonging to one set or another to a degree. Multiple data points could help indicate to what extent a given item belongs to a given fuzzy set.

This technology is used by some car manufacturers in their automatic transmissions to determine optimal shift points.

At this point in time fuzzy logic had not been applied in software for blocking spam. However, to me it seemed like an ideal system for building a better anti-spam filter.

Next...