I came to think about how spam-filters on mail clients work, and most of them use bayesian filters to catch the trend of spam an account receives (or maybe the opposite, a filter of whats not spam…)

So, would it be possible to teach the filter not to receive certain legitimate emails?

How? By spamming of course! Bayesian filters filters would quickly learn if such and such words appears in spam. Think of it as a Denial of Service on legitimate emails. Name your least favorite politican and add a good offer too good to be true (use the signature maybe?)

Any comments on this idea, am I hitting a dead end or could this actually be possible?