Kees Theunissen wrote: > [snip] > >>There is a small problem with this approach - Bayes database do not >>learn phrases and words used in e-mail sent by your own users. >> >> > >Is that a problem if you don't scan these messages anyway? > >
That's a bonus, if you ask me. If you post to a mailing list with a lot of traffic, like alsa-users for instance, someone could use the text of your own postings (if they wanted to work hard enough) to make the spam look more legitimate (or at the very least, they could do a Markovian chain analysis of the list to see what words in that corpus are specific to it. You could see people (harvesters) selling not only lists of addresses, but also "magic pass phrases" to lower the defenses of Bayesian filters. It's like the grifter trick of listening to two blokes in a bar talking about a third person, and then later approaching one of them at telling him you know "Joe" (or whomever) to get into his confidence.. -Philip _______________________________________________ NOTE: If there is a disclaimer or other legal boilerplate in the above message, it is NULL AND VOID. You may ignore it. Visit http://www.mimedefang.org and http://www.roaringpenguin.com MIMEDefang mailing list [email protected] http://lists.roaringpenguin.com/mailman/listinfo/mimedefang
