[H-GEN] The declining effectiveness of SpamAssassin

[Michael Anthon]

> began using it, I'm still using version 2.44.  I'm planning to
> test version 2.63 in the next few days to see whether it has
> evolved enough to make it more useful.

Well, yes, the later version will be heaps better of course, but I'm
finding a few things getting through even in the lst few days.
Generally stuff where all the keywords that would normally trigger some
SA rules are misspelt (typically spams for vigra...).  No doubt I should
teach these to the bayesian filter to help it out a bit.  

Another thing I have found to be very useful in reducing the false
negatives is to look at the rules that the message in question triggers
and investigate changing the score from the default.  In particular I
have found increasing the scores for some of the block lists to be
highly effective.

To this end I have the following in my /etc/spamassassin/local.cf...

score RCVD_IN_NJABL 3.0
score RCVD_IN_UNCONFIRMED_DSBL 3.0
score REFERENCES 0
score MICROSOFT_EXECUTABLE 20.0
score RCVD_IN_MAPS_RBL 3.0
score RCVD_IN_SBL 3.0
score RCVD_IN_OPM 3.0
score RCVD_IN_SORBS 3.0


Cheers,
Michael