In my environment, I have thousands of users, who often don’t
bother to read the instructions on where to send false positives and such, thus
they either do not report spam/nonspam, or send it to the wrong place.
I have done the following, not to correct their actions, but
to make sure ASSP gets taught about most things it misses:
1: enable IMF in ex2k3.
2: download and use IMF archive manager to view the imf
3: in the IMF manager, configured the reports to be from an
account where you want to see verification of the email interface submission,
and the TO: to be your assp ‘thisisspam’
email interface. In my case:
Since I am in a training faze, IMF manager shows dozens of
spam every few min, with usually about 10 missed by ASSP per every 10 min. I
sort by subject, then select the items that ASSP missed and are obviously spam,
and hit click ‘Report’ a bunch of times in hopes of teaching ASSP
quickly. While this obviously does not get the things that both ASSP and
IMF misses, it at least assures that as much can get reported as possible when
you have super-unsavvy users;)
It seems to have drastically increased ASSP’s Bayesian
accuracy in the past few days.
Once out of a testing faze, I’m hoping ASSP will have
the ability to tag the subject on all spam CC’s, so that I can continue
to teach the filter on occasion.