John Thompson writes: > On 2006-10-16, Marc Perkel <[EMAIL PROTECTED]> wrote: > > > What need to be done with messages that are spam is to only learn the > > headers and not the body of the message. What needs to be done is some > > detection of deliberate bayes poisoning and removal of the poison before > > larning. > > Does Bayes poisoning actually work, though? C.f.: > > http://www.zdziarski.com/papers/boudville.txt
alternatively: http://www.cs.dal.ca/research/techreports/2004/CS-2004-06.shtml That's only certain classes of Bayes filters, however, and (if I recall correctly) it looked likely that SpamAssassin's use of the Robinson/Fisher chi-squared combining algorithm defends against it quite well. But in my opinion, Bayes poisoning can definitely work, based on the use of "common words" rather than simply "random words". --j.
