SpamBurglar was written by Tom Ashley, Matt Hansen, Marie Joiner, Jen
Knutson and Josh Ourisman as our senior project for Carleton College's
"Integrative Exercise". (a.k.a. Comps) The advisor for this project was
Dave Musicant, Professor of Computer Science at Carleton College. The
Bayesian filtering algorithm used in this system was derived from a paper
written by Paul Graham, which can be found at
http://www.paulgraham.com/spam.html.
How to use the program:
When you run SpamBurglar, a window pops up with three options: Proxy
startup, Configuration and About SpamBurglar. The contents of each window
is as follows:
Proxy startup: Starts and stops the local proxy server. The proxy server
must be started for the filter to operate. The proxy server must be
stopped before the program is exited.
Configuration: Sets up the various user preferences needed in the
operation of the program. The "Host" field indicates the mail server you
wish to filter spam through. The "Port" field indicates the port number
for this server. The "Proxy Port" field indicates the port number for the
local proxy server which is responsible for the actual spam filtering.
The user must select the correct protocol used by their mail server,
either POP or IMAP. The user must also determine the action the filter
should take on messages identified as spam. The filter will either mark
the message as spam by appending a notification to the subject line, or it
will delete the message entirely.
The user may specify the "threshhold"
probability for determining if a message is spam. The default is 90%.
The window also has an option to edit a "whitelist". The whitelist is a
list
of email addresses that will always be considered to be senders of
nonspam. The user must enter the entire address: username@host. If a
user wishes to add all users from a particular host, the syntax is
*@hostname.
Finally, the user may choose to enable or disable "Training mode". When
training mode is enabled, a window will appear at the end of the
SpamBurglar session with a list of all new email messages and their spam
identities. The user has the opportunity to correct/change the results if
necessary, and the program will use this new data to update the filter's
table of spam probabilities. Training mode should always be enabled
unless you are sure that the filter has been adequately trained on your
email.
Note: SpamBurglar is licensed under the GNU Public License.
Joshua Ourisman
Last modified: Thu May 13 11:34:10 CDT 2004