Going back to my concept of the container of information for a moment, a
watermark -- in the real world -- is a characteristic of a document's
container, ie. paper, not the information in the document.
You are, of course, correct. There is a longstanding tradition of changing
certain data within a document to identify its source but this is usually
only used to prove(?) that published data has been copied. It does not
identify the copying agent nor the distribution path the copied document
has taken. There are only so many things a publisher can do (or is
willing to do) using this technique simply because it obscures the value
of the document.
Watermarking information, if it were possible, would allow each copy of a
document to have its own identifying mark, tuned to the original
recipient, thus providing proof that a certain instance of a document
which was provided to a certain person, was then distributed further.
Using the technique of injecting error into a document to identify its
origin is only as useful as the limitations of the tools we use to
manipulate it. Many of the tools we currently have empower the user to
make their own presentation decisions -- discarding multiple spaces within
a paragraph, for instance. While this has not yet been used to change
the data once it has been retrieved -- perhaps for efficiency sake -- I
would not consider it completely out of the question.
The only kinds of error we could, I believe, use to track the origination
of a document would be visible enough to affect the reputation of the
publisher. I know of no successful publisher that would knowingly
distribute documents with spelling or punctuation errors.
</rr>
On Sun, 18 Sep 1994, Karl Auerbach wrote:
>
> > And I thought I had been very clear in stating that you cannot watermark
> > information.
>
> It's apparently a fairly common practice to do so. The examples I have
> heard are mapmapkers who put in false towns or slightly change the
> height of mountains. That way copying becomes pretty obvious.
>
> In software, I put in little bugs (ones that aren't normally triggered.)
>
> Yes, it does make the "information" a bit less accurate, but not
> significantly less so.
>
> --karl--
>