I know we would all love to have a tool that automatically knows what is spam and what is not spam without any work from us, but I have noticed the active war in progress between the spam blockers and the spam generators. The ingenuity of the spammers challenges all spam filters, and leaves us settling for minimizing the time we need to spend in training our own spam filters.
I like SpamBayes because I am more certain that it is not classifying good email as certain spam. A little spam is classified as good email, but much less than my actual good email. I have now trained my SpamBayes so I hardly ever see a good email in my possible spam folder, so I am quite certain that no good email is going into my certain spam folder. In order to get to this level, I have trained over 3000 spam and 3000 ham messages. Since my initial training was nearly this large, the amount of additional ham and spam training since initial training has only been in the hundreds (very little ham training). I have achieved this with my filtering of certain spam at 75%, and possible spam at 15%. The fundamental work we must do to train SpamBayes can fortunately be focussed on the easier task of training SpamBayes on the contents of our possible spam folder, and then occasionally telling SpamBayes that something it thought was perfectly good was actually spam. I find processing the possible spam folder is much faster than if this email is in my inbox. I always process my possible spam folder before I look at my inbox. Although you can imagine rules that would reduce the requirement to train SpamBayes, the engineers at SpamBayes have determined that none of these algorithms work. All algorithms that identify ham without using full SpamBayes will allow spam to be classified as ham. Only SpamBayes approach can actually protect you from the spam, but it must be trained, and re-trained as the spammers alter the nature of their attacks. You can see the sophistication of the spammers you are dealing with and SpamBayes in defending you by looking at the "Show spam clues for current message" button in Outlook. This writes an email for one message showing you the entire SpamBayes analysis of this message given your training database. As I look at this, I am astounded by which tokens actually turn out to be useful in separating the ham from the spam. SpamBayes is able to find and use tokens most of us would not even think of using. Sometimes we wonder if it isn't looking at too many tokens, but sometimes tokens we wouldn't think would be useful for separating ham from spam do turn out to be useful, after all. It turns out that the token "cc:" is a very useful ham indicator for me, with a spam probability of 0.00614804, whereas no-one would expect this, but SpamBayes is using it in my database. Spammers could eliminate the usefulness of this token at any moment. As soon as they do, SpamBayes will immediately adapt. Peter Bishop Aeroprise, Inc. Take advantage of the Aeroprise Enterprise Discovery and Personalization System for both Smart Clients and standard browsers available only with the Aeroprise Mobile Gateway. -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Amedee Van Gasse Sent: Friday, December 01, 2006 11:12 PM To: David and Maureen Jardine Cc: [email protected] Subject: Re: [Spambayes] Can I configure Spambayes to accept all mailfrom trusted addresses? Op zaterdag 02-12-2006 om 10:35 uur [tijdzone +0800], schreef David and Maureen Jardine: > Spambayes keeps putting mail from friends and even my own work email > address into the Junk Suspects folder. Is there any way that I can > configure Spambayes to accept all mail from particular email > addresses? Yes, you can train them as ham. See also the thread that started on 21/11 about the same topic. -- Amedee _______________________________________________ [email protected] http://mail.python.org/mailman/listinfo/spambayes Check the FAQ before asking: http://spambayes.sf.net/faq.html _______________________________________________ [email protected] http://mail.python.org/mailman/listinfo/spambayes Check the FAQ before asking: http://spambayes.sf.net/faq.html
