From: "Perry E. Metzger" <perry(_at_)piermont(_dot_)com>
...
I use a very large variety of techniques to block spam, and I'm
something like 95% successful. The 5% is starting to kill me, and
making things substantially more successful than that is likely not
possible without blocking lots of legitimate mail.
...
If you're are getting only 95% with a large variety of filters,
then you should throw them all out and start over. Two independent
80% filters should be good for 96%.
There is no single silver bullet for spam, but there are things that
help a lot. There are many tactics that do better than 5% false
negatives (i.e. filter 95% of spam), with varying false positives for
various individual situations. My personal combination of filters
(which would almost certainly *not* be appropriate for your situation)
does better than 99% (i.e. fewer than 1% false negatives) with fewer
than 1 false positive per month (i.e. fewer than 0.1% false positives).
Perhaps the most important aspect of spam filtering is it needs minimal
participation from the mailbox owner. To do better than about 60%
false negatives with fewer then 10% false positives, the mailbox owner
must at least be willing to maintain personal white and/or blacklists.
This fact is sort of the obverse of another fact, that if more than
1% of spam targets complained about spam to ISPs (as opposed to the
fewer than 0.01% in practice), then spam would never have been a
problem.
Vernon Schryver vjs(_at_)rhyolite(_dot_)com