Bayesian Filters

Spammers are now targetting our Bayesian filters. I've been doing some research on it, but there isn't a huge amount that we can do. The only feasible option is to work on implementing more custom rules for the main spam filters and see if that has any effect. Over the last few days there has been a marked increase in the number of emails that are getting through. Fortunately most of the mails that get through do not contain anything offensive, but they could have done! This is the real problem. What would have happened if the more offensive types of spam mail had reached one of our users? Luckily the RBLs seem to be working well.