One thing that I use for combatting Spam is a bayesian classifier. Basically, it takes Spam that you recieve and trains the classifier to know when you are receiving Spam that meets your criteria as being Spam. A couple of people on IRC have asked for my Spam archive to train their classifiers. Here is my human-verified spam archive in bzipped mbox format containing 754 messages. There’s another archive of 458 messages here.