My undergrad thesis, written four years ago, was a taxonomy of anti-spam systems. Since then, every anti-spam system that I've seen both (1) falls neatly within one of my categories and (2) fails to account for the problems I identified in that category.

This idea is an authentication-based system, because it relies on identifying the message sender in order to prevent spam. Among other things, it fails to account for identity theft.

For reference, here is my thesis: http://www.eecs.harvard.edu/~cduan/download/?file=thesis.pdf