Where can I find a Bayesian spam filter that will run on our Windows and Linux workstations?
POPFile is a free Naive-Bayesian spam filter available at popfile.sourceforge.net. It is written in Perl to work with POP3 mail clients, and runs as a POP proxy on your workstation.
After installing the software you change the POP server settings in your mail client to use the POPFile proxy, which then handles the communications with your POP server. No changes to the outbound Simple Mail Transfer Protocol server settings are required.
The Windows version comes with everything it needs to run bundled in a Windows installation package, and instructions for using it with Outlook and Exchange Server are provided.
A Web browser interface lets you configure, manage and train the Bayesian mail filter. POPFile supports multiple accounts, and it can be used with multiple e-mail clients on one machine or on a gateway server for a network. Mail also can be read through the Popfile browser interface.
When you install POPFile, you teach it the difference between the mail categories you establish, called buckets, by classifying a number of messages by hand through the browser interface. After that, you only have to reclassify errors. Initially, I hand-classified 40 messages, and have had to reclassify less than a dozen out of more than 1,000 messages since.
