Slashdot Mirror
Startup Tries Watermarking Instead of DRM
Loosehead Prop writes "A U.K. startup called Streamburst has a novel idea: selling downloadable video with watermarks instead of DRM. The system works by adding a 5-second intro to each download that shows the name of the person who bought the movie along with something like a watermark: 'it's not technically a watermark in the usual sense of that term, but the encoding process does strip out a unique series of bits from the file. The missing information is a minuscule portion of the overall file that does not affect video quality, according to Bjarnason, but does allow the company to discover who purchased a particular file.' The goal is to 'make people accountable for their actions without artificially restricting those actions.'"
It takes a man to suffer ignorance and smile
Be yourself no matter what they say
ssand,
Did you read even the summary? Removing the beginning does not remove the unique signature formed by bit removal.
Of course bit removal or any sort of water mark can also be mucked with.
Still, this would be more user friendly than "hard" DRM.
In principle, I like this idea. I don't really see a problem with it.
However, they already do something similar in theaters. Every so often in theatrical movies you will notice a weird pattern of "cigarette burns" that appears for a brief moment. (Yes, to my eyes at least, they are visible and sort of distracting.) The pattern is different for each copy of the film shipped. The idea is that, if someone sneaks into a movie theater and makes a cam of a first-run movie, the producers of the movie can analyze the video and figure out which theater it came from. That helps them put more pressure on theater owners to enforce bans on video cameras, etc.
But does it seem like there are fewer cam bootlegs out there since they started doing this? They started it maybe five years ago.
Breakfast served all day!
The "watermark" is throughout the video. The first five seconds is just a "header" if you will, for the naked eye to see. The watermark however, could not be removed so easily.
Starmen.net
I take it you didn't even read the summary, let alone the article. Hint: your solution won't work.
from TFA-
"Because of its design, the watermark even survives most editing changes and format shifts"
A good watermarking solution is resistant to many kinds of processing, including re-encoding.
Thats the whole challenge.
Off course the watermark might not be resistant to extremely destructive transformation such as downscaling from HD to QCIF, but then who cares about pirated QCIF video ? But certainly a very accurate transcoding would not affect the watermark.
Current watermarking technologies are very much dependant on proprietary algorithm.
"and the people I know who buy digital download media, don't use illegal filesharing sites"
I don't know about you, but I know plenty of people who go to such filesharing sites because they are unable to obtain the media in a usable format if they pay for it. If they can't view the "legal" media for whatever reason (unsupported mobile device, Linux user, etc), then the legal media becomes worth $0.00 to them, and they go the illegal route.
Now, if the legal media were usable to them, then it would actually have some value to them. Why pay for something when you can get it for free? Convenience and selection. Back in the days of pyMusique I bought more music in a week than I did in 3-4 YEARS, none of which has ever been copied to a device I don't personally own. Same with allofmp3. While allofmp3 had its own issues in that the money flow from the user to allofmp3 broke down there rather than flowing to the RIAA, the RIAA were incredibly stupid not to look at allofmp3's success as proof that people WILL pay for music they can get for free if you offer them convenience and compatibility. If you can't beat em', join em'. The RIAA could have made a killing by creating a site with:
Better selection than allofmp3 (somewhat difficult, but not really for the copyright holders themselves)
Better convenience than allofmp3 (accept more credit cards, and simply be a "trusted" vendor people are comfortable providing CC information to)
Double the price of allofmp3 (still quite reasonable)
No DRM
retrorocket.o not found, launch anyway?
EZTakes is already doing the same thing in the US. A brief intro before the film shows who purchased it. I've tried ripping the disc just using selected titles (as the "watermark" intro is just a title on the DVD) to remove my name, but they've anticipated that also. I haven't got a good rip yet without the watermarked intro.
Not that I wanted to rip them off or anything. I just took it as more of a technical challenge as to whether I could remove the watermark on the otherwise non-DRM'ed DVD.
Pretty cool service, although you won't find any blockbusters in their selections quite yet. But for my taste in '70s cult films, I've found a few that I've downloaded for the fair price of $2.99. Not bad, not bad at all.
:q!
I've been very careful but 8 people use my computer, I get viruses way to often and at least three times I was part of botnets without my knowledge."
You've gotta be kidding - at that level of infection the appropriate phrase is "I'm very ignorant".
If you're assuming the bits will be in the same place in each file, and it's just a trivial case of doing a diff. you're very naïve. Digital watermarking schemes are generally quite advanced these days, encode the data in many redundant ways thoughout the file, add in chaff with the wheat to foil attackers, and are resistant to many transformations (scaling, transcoding, etc.). That's not to say the watermark can't be removed, or at least corrupted. See Felten's SMDI research for a perfect example. But it's far from trivial to do so with a well designed watermark.
"The invisible and the non-existent look very much alike." -- Delos B. McKown
The technique from the article is called "fingerprinting" in the watermark literature, and comparing 2 copies is known as a "collusion attack". Existing fingerprinting codes are quite resilient to collusion attacks from two colluders (the code construction and proof of performance requires a good grasp of information theory concepts, so I'll skip the details).
As the number of colluders grows, however, the reliability of present codes drops drastically. In addition, I don't know if analysis using fingerprinting codes would be admissible in court: the best results simply show that if n fingerprinted copies are used in a collusion attack, then with some probability (ideally approaching one, but not there yet with present codes) at least one of the n colluders can be identified from the modified copy. Present values of n tend to be small.
Whether this person knowingly contributed to the attack is a separate question. It's easy to devise scenarios where one of the "colluders" was in fact played for a fool or fall guy, which is why one can cast serious doubt on any forensic efficiency of present-day fingerprinting codes.