Since spam-sending accounts have names that are clearly computer-generated, I have been wondering if an algorithm can detect the most obvious cases.  I have an idea that I am going to test, but I wondered: Has anyone else has made a similar attempt?  If so, what failed a d does anything work?