I'm working on a server that has spamassassin processing messages that have already been marked as spam by a primary filter on another server. The reason Iím doing this is because there has been some complaints about people not receiving some subscriptions and other emails that have been caught by the primary filter's bayes filter. Someone told me of an idea to set up another spam filter to evaluate the spam a second time. In theory, all messages marked as spam by both the primary filter and spamassassin would most definitely be spam then, would be deleted without further review. Messages marked as spam by the primary filter but not spamassassin would be marked as uncertain then those messages would be evaluated manually. This server with spamassassin receives about 20000 emails a day that have been evaluated as spam by the primary filter.

The problem I'm having is trying to teach spamassassin's bayes filter. I have plenty of spam messages, but I don't have enough ham messages. Does anyone have any ideas as to remedy the situation, or perhaps some alternatives to a secondary filter? Thanks.