> From: Gary Hughes > Sent: Friday, April 02, 2004 11:58 AM > > I've noticed a lot of the junk mail now includes a lot of random > words, presumably to fool systems that use Bayseian analysis. > Is it coping with those? Yes, these don't fool SpamBayes. I think this is due the fact that SpamBayes includes some of the message headers in it's analysis, and that most of my spam is HTML while most of my ham (ham is good email) is not. HTML tags make for great filters in my case! The beauty of Bayesian filters is that each user builds a database specific for their email traffic. With only a 100 or so messages, SpamBayes was over 90% accurate for me. Now, with 682 spam 202 ham messages in my training database, it is performing very well (greater than 95%, maybe even above 99%). I could have trained it on 1000's of spams, but it is working well with this small training set. I now *never* have spam in my inbox. It does put a few messages a day into the "suspect folder." Most are spam. Every once in a while, a ham email gets into the suspect folder. I haven't had ham in the "spam" folder since the first days of training SpamBayes. I currently get about 100 spams a day. As I said before, I can't say enough good things about SpamBayes. I have to be careful, however, otherwise I'll sound like one of those spam-blocker spam emails! I am really nothing more than a very satisfied users of this free, open source software. If you have problems with spam, you owe it to yourself to check out SpamBayes. (By the way, if you want a supported commercial product, check out the "related" section on the SpamBayes sourceforge web page, http://spambayes.sourceforge.net/) -Gary Segal ---------------------------------------------------------------------- You can UNSUBSCRIBE from the OpenDTV list in two ways: - Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org - By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word unsubscribe in the subject line.