Tatsumi commented the other day that she was a little disappointed (to say the least) with our current anti-spam software. The stuff we are using now is part of the NIS 2003 Pro software and while it did an OK job (about 60% success) it just wasn’t doing as well as we hoped.
Interestingly, just after this I came across a thread/discussion on Slashdot about this very topic, so I read through it looking for alternatives. I had a short list of requirements and preferences.
- WindowsXP compatible
- Easy to install
- Usable by Tatsumi and Kimiko
- No conflict with or existing spam, firewall and anti-virus software
- Free
A lot of software showed up in the thread, most of it centered around the concept of Bayesian Filtering. Most notably as discussed in a paper by Paul Graham. You can read below to see what I wound up with as well as a lot of info on the topic.
I’ll let you know how it goes later on 🙂
What I picked:
In the end, I wound up working with SpamBayes in the form of the Outlook plug-in partially because of the review Jon Udell gave the combo. It will take me a little while to train it, but it sure looks like a good start.
Resources:
- Slashdot Bayesian Filtering For Dummies – a nice thread with some good info.
- Spambayes Mailing List – this list has emerged as one of the central places to discuss spam filtering.
- Gary Robinson’s Spam Rants – a weblog devoted to stopping spam.
- wecanstopspam.org – a resource page on stopping spam
References:
- BBC NEWS Technology How to spot and stop spam – an article discussing methods of spam detection.
- A Plan for Spam – the August 2002 paper by Paul Graham laying out the basic concept.
- Better Bayesian Filtering – the January 2003 refinement by Paul Graham.
- Spam Detection – a long article by Gary Robinson that gives some good info and puts some of Paul Graham’s work under a microscope.
- SpamBayes Background Reading
Seriously technical references:
- A Tutorial on Learning Bayesian Networks – a very technical paper defining true [[wp:Bayesian]] networks.
- [cs-0006013] An evaluation of Naive Bayesian anti-spam filtering