A while ago I moaned about some eegit who has been using my domain name to send Unsolicited Commercial Email (aka Spam) to the great unwashed.
The problem as it now stands is that 90% of my incoming emails are bounce-back messages from servers, claiming messages I've sent cannot reach their intended recipients. Of course, I never sent them in the first place.
I collect all the mail sent to my domain, and those emails that are not addressed to specific user accounts are fed into SpamAssassin to help its Bayesian filter. However, this means that a large number of bounce back messages are also being learned.
What I need to find is a realiable way of determining whether a message is a rejection notice or not; I could then insert this into a maildrop filter so that only the original spam messages enter SpamAssassin's corpus. Anyone have any ideas on how to achieve this?