How do you prove a computer-generated work was independently generated by you?

This question is related to one of my earlier posts about the ability to copyright sudokus. I have written a program that generates sudokus and I want to make a website with these sudokus. It seems that it is up for debate if computer-generated works (like sudokus) can have copyrights on them, however it was suggested that it would be in my best interest to keep track of proof that I generated these works on my own. Does anyone have suggestions on how to keep proof that you independently generated a given computer-generated work? How extensive does the proof need to be?

First off, you are highly unlikely to ever be challenged on this. But if you want to, it's not that hard to secure a chain of evidence.

1. Most important: Use a version control system (Git, Mercurial, Subversion etc) for your codebase. The record of version control checkins is great evidence to show in court. It shows a) who built the codebase (the authenticated users associated with each checkin), and b) the time, how your codebase developed over time, and which features landed when. For bonus points, use a version control system hosted by a 3rd party so you can't manipulate timestamps etc -- f.x. Fogbuz, Github, UnFuddle etc.
2. Have some basic build instructions, and tag release branches clearly. That way an impartial 3rd party analyst can build the software from your SVN source tree, and verify that the resulting compiled binary is bit-for-bit identical with the running version on your servers. This provides evidence that your source control tree is complete.
3. Save a log file of the seed parameters for your "computer-generated work" matched to the ID of the work. That way you can prove that your binary, given these seed parameters, actually produces the content seen on your site.

If you ever go to court (again, you almost certainly won't), then one of the first steps will be to find a qualified technical expert to go through the evidence, and explain it to the judge. In my experience, the best local universities are good (but expensive) place to find such experts -- the academic staff has a high social standing and reputation for scientific objectivity.

Here is the nerd way to prove your timestamp: tarball your program, take a SHA512 hash of the tarball, then publish that publically in a place that is essentially permanent (classified section of the New York Times, eg).

Now any change you make to the code will upset the hash value, so no one can claim that you have modified your program since the time of the hash publication.

And it's very cheap.

"having a dated evidence trail is what matters"

Yes, but a lawyer isn't a 'recorder of authorship'. Stuff like 'poor man's copyright' (sending yourself a registered letter and not opening it) doesn't work because it comes down to one's word which the court would of course not accept (the instant one signs for the letter it's no longer assumed sealed by a third party).

A provisional patent lets the gov't register the info but it's $375? so it might not be a good option should you want to iteratively store the code.