Amazon Patents Changing Authors' Words 323
theodp writes "To exist or not to exist: that is the query. That's what the famous Hamlet soliloquy might look like if subjected to Amazon's newly-patented System and Method for Marking Content, which calls for 'programmatically substituting synonyms into distributed text content,' including 'books, short stories, product reviews, book or movie reviews, news articles, editorial articles, technical papers, scholastic papers, and so on' in an effort to uniquely identify customers who redistribute material. In its description of the 'invention,' Amazon also touts the use of 'alternative misspellings for selected words' as a way to provide 'evidence of copyright infringement in a legal action.' After all, anti-piracy measures should trump kids' ability to spell correctly, shouldn't they?"
canary trap (Score:3, Informative)
Re:Quick! Someone say it's only defensive! (Score:2, Informative)
They'd NEVER file multiple lawsuits against people for infringing totally obvious patents, right? Of course not! That'd be like saying that Slashdotters actually believed half the stuff they said about freedom and rights.
Quick! No one's said anything stupid yet! Let's construct a straw man so I have something to ridicule!
Canary trap (Score:5, Informative)
Intelligence agencies have been doing this sort of thing for decades, giving slightly different versions of a sensitive document to suspected spies or places where possible spies might have access to it, with some subtle changes in the words, seeing which one gets leaked or appears elsewhere. Tom Clancy coined the term Canary trap [wikipedia.org] for the technique. Patriot Games was published in 1987, but its real-world use for exposing information leaks most likely predates the novel.
Moral rights (Score:4, Informative)
Canada and some other countries have "moral rights" which belong to the author.
Changing words without his permission could violate these rights.
In some countries these rights are inalienable and non-assignable. This means the author can't be ordered to waive them by the publisher or other copyright-holder.
Re:Canary trap (Score:5, Informative)
Intelligence agencies have been doing this sort of thing for decades, giving slightly different versions of a sensitive document to suspected spies or places where possible spies might have access to it, with some subtle changes in the words, seeing which one gets leaked or appears elsewhere. Tom Clancy coined the term Canary trap [wikipedia.org] for the technique. Patriot Games was published in 1987, but its real-world use for exposing information leaks most likely predates the novel.
But the classic Canary Trap requires someone to modify the document manually, which is hard to do on a large scale. Here it is being done automatically by an algorithm.
However, I am aware of published methods for this problem dating back to 2001 [trnmag.com] by Mikhail Atallah at Purdue. In fact Atallah received a patent for followup work [uspto.gov] in 2007, a year after the Amazon patent was filed.
Here are a few hundred papers [snipurl.com] on the subject, via Google Scholar. Some adjust whitespace, some modify images of the text, and some attempt fairly sophisticated syntactic analysis and restructuring of selected sentences.
I apologize that I haven't read the Amazon patent, or read the prior literature carefully, or gone to law school, so I can't comment on whether the patent seems valid or not.
Re:Prior art (Score:4, Informative)
... the word "aluminum". Some use the US spelling, the other use the correct spelling.
FTFY
Re:Patentable? (Score:5, Informative)
I don't blame people for not reading the claims section, because it's necessarily an obtuse fusion of legalese and jargon.
But no, they did not patent *doing* this, they patented the *way* that they do this. Patents cover implementations, and not ideas. Some have argued that the line has been blurred with certain classes of patents, but it hasn't blurred so far that the concept in the slashdot summary is actually locked up as IP.
Frankly, I can't be bothered to look at the claims either. But the idea itself certainly lends itself to ideas that are patentable (whether they should be patentable or will be rendered retroactively invalid is another question). For instance, I'm curious how they identify which words should be replaced, and the system by which they choose a synonym that hopefully doesn't destroy rhyming patterns, metrical rhythm, puns, shades of meaning, and ambiguity in words with multiple meanings that don't completely intersect the candidate synonym's meaning.
Also, whatever they are they doing to prevent the trivial case of three copies being compared to recover the original. Maybe they have a bunch of sets of synonyms that are commonly replaced so you need more to get the original, but even then, do they arrange it in some way so that the source of the leaks can be traced down despite the alteration? Or maybe they just assume that book pirates are morons.
They might do nothing for any of those cases, mind you. Once again, I can't be bothered to read these damned things. Which is part of why I don't submit articles about ones that I've decided I think are actually stupid.
Re:Patentable? (Score:5, Informative)
Re:Moral rights (Score:2, Informative)
Re:It's not about the patent, it's about the lying (Score:1, Informative)
Yup, I'm (ahem) fairly knowledgeable about this field, and this is a really dumb (and elementary) scheme. They shouldn't be granted a patent on it, as it's a textbook example of what not to do in watermarking.
Why? Because it's noticeable, and they've just told everyone about it. Given that this is a field overwhelmingly reliant on obscurity, that was really not a very smart thing for them to do.
The goal of steganography - and watermarking is a form of steganography - is to remain undetected to an attacker. The most powerful thing a watermark can be is invisible, because it's only got an advantage when the attacker doesn't know it's there. When an attacker knows that a watermark is present, they can use an attack such as a collusion attack (which you described, and which is one of the most powerful classes of steganalysis/antisteganography), although other powerful attacks also exist.
While some schemes attempt to be resistant to this, nothing (even a mark tree) is actually immune, especially if you know how the marking scheme works. It's not even true that the stronger and more noticeable you make a stegomark, the harder it is to remove. Sometimes very strong ones are easy to remove (although weak ones are never hard). And the outcome from a successful colluding removal will be higher-quality and more compressible than all the watermarked copies, because you've removed the (deliberate) noise.
Unfortunately to the would-be watermarker, as a matter of theory, optimal lossless compression provides a perfect stegomark detection, and optimal lossy compression provides a perfect stegomark removal. Most of the compression and filtering techniques out there make absolute mincemeat out of the entire watermarking field, providing for very powerful detection and removal attacks, and they can even improve the quality on the way (in this field, for example, instead of a scanned PDF containing the encoded noise of a watermark, even scattered in written form - which will reduce the compressibility of the text as introducing a stegomark raises the entropy - the attacker could collude amongst 10 or 15 different copies and reconstruct the source text, producing a smaller, completely clean PDF with no unnecessary scanned images or noise; a file of identical or superior perceived quality in a smaller form is generally more desirable).
A more subtle technique of tweaking the typesetting - inter-word tracking, perhaps, or minor changes in line spacing - would be more effective because it is more likely to go unnoticed, but once the attacker knows a watermark may be present, the game is up.
And that's what makes this scheme really dumb; that they've told everyone about it now, so now people know to look for it. Besides, I can't imagine the authors being particularly happy. Their publishers may have rights to edit for copy-setting or localisation purposes, but the authors may retain rights of approval in some cases, and I'm pretty sure that right would never extend to the retailers...