It sounds like she's just a tad idealistic about the security and benefits of her particular scheme But it does illustrate a method that is used as part of more complex schemes.

I disagree with her that the code ought to be able to check its own watermark. While that sounds like a good idea in principle, I think it would backfire in practice. If the code reads its own watermark, you've just given the reverser all the information they need to both find and decipher your watermarks. The whole idea of this kind of watermark is to remain completely undetected. So I would think that your watermarked code should avoid interacting with the watermark entirely.

Here is a good (but much more difficult to understand) thesis on actual code watermarking techniques:
http://www.julienstern.org/files/codemark/
The bibliography gives a lot of good reading material, too. You might find it useful to peruse his site in general - there's a lot of good reading there.

I get the feeling that it might help to read "A secure, robust watermark for multimedia" from Lecture Notes In Computer Science, Vol. 1174 before reading Stern's thesis, but I'm having difficulty locating the actual text.

Ok, I did finally find it, but it's terrible quality. Many of the equations are nearly unreadable, unfortunately:
http://www.cs.bgu.ac.il/~ns041/cox_et_al.pdf

Here's what looks like a later revision of the paper, but it's also in a really bad state:
http://www.assuredigit.com/tech_doc/more/IP97.pdf

[EDIT:JMI Please notice the "Edit" button was designed to make additions to a post. A "new" one is not necessary.]