The junk E-mail plague just seems to get worse and
worse. Annoyance Filter is a tool which
permits individuals to immunise themselves against
time-wasting and offensive unsolicited E-mail. It doesn't
kill the rats or fleas, but at least it keeps the pathogen
they propagate from ever affronting your eyes.
Annoyance Filter is an adaptive,
trainable E-mail filtering program. You train it
to distinguish mail you're interested in reading from
junk you don't wish to receive by supplying it with
a collection of each kind of mail—the more the better.
Then, when a new message arrives, the content of each is
compared with statistical profiles of legitimate mail and
junk computed from the training collections, and the
probability the message's being junk is estimated. If the
junk probability exceeds a given threshold, you may wish
to automatically discard the message or quarantine it in
a containment facility for later review. Junk mail evolves
“protective colouration” over time to evade conventional
filters, but this is of no avail against
Annoyance Filter, which can be re-trained
at any time with contemporary mail and junk collections.
Annoyance Filter is written in standard C++ and can be
installed on any system with a suitable compiler. On Unix systems,
Annoyance Filter can work in conjunction with
Procmail. There is nothing Unix- or Procmail- specific about
Annoyance Filter; it can be integrated into any mail
processing system which permits an external program to filter
incoming mail. Developers familiar with Windows, Macintosh, and
other proprietary mail clients are encouraged to adapt the program
for use with them and contribute their work to this effort.
New release 1.0d builds with GCC 3.4 as well as earlier
versions, and adds the ability to parse deliberately mis-formatted
MIME headers some junk mail uses to avoid filters.