I strongly recommend you read Richard Clayton's paper on how (among other things) one could hack the Cleanfeed system to *find* the really bad stuff. He and his colleagues at the Cambridge Computer Lab also have a fine blog - http://www.lightbluetouchpaper.org