Bayesian spam filtering is an incredibly powerful statistical technique—with acceptable computational complexity—for identifying spam messages. Bayesian techniques address many weaknesses of other methodologies:
The entire message can be examined, not just special parts.
All words are significant, not just special keywords or addresses.
Updating is, in practice, infrequent (never more than one or two email messages per week through the training program; often none).
So far, spam attacks on Bayesian filters have been relatively unsuccessful.
When combined with other techniques, Bayesian filters can be a very strong component of an institution’s global spam system.
The CRM114 system is just one of many available Bayesian spam filters; commercial ...