NEC announces video checking technology

The criticisms of this technology are based on a misunderstanding of how it works. It works by creating a "checksum" that reflects roughly what the image looks like, similar to how Tin Eye compares images.

It compares sequences of frames against a library of such checksums and can identify sequences of images that correspond to any work in its corpus. You do not need the original work to find a match, just as you don't need a copy of every virus for your virus scanner to match them.

Small alterations such as single pixels or the like won't work. The checksums will still match. They compare what the image "looks like" (as a human would perceive) it.

You can think of it as something like "large round blob on the left, moves to the right as a small square moves off the right edge". This is not literally what it is, but it is a similar idea.

Of course, it cannot determine if use is authorized or unauthorized or whether unauthorized usage is fair use. However, this is really not that hard to work around.

First, you create a corpus of only works who distribution is not authorized, period. Then you set your match threshold to require a match of over a particular threshold for a particular amount of time.

Of course, you still have to decide what to do when you find a match. One possibility is to notify the uploader to confirm that they still wish to upload and tell them the copyright holder will be notified. Then notify the copyright holder, and they can decide whether they have legal grounds to send a takedown notice.

The first thing the criminal will do to ID the drm is look for it. That means it will know they use 60frames and search for a pattern in 60 frames and compare it to a database even if they have to use their eyes to do it.

So, to get rid of this you must confuse their search, and you do this by staggering the drm ID checksum. Use multiple checksums for a entire move, a different one each 60 frames, and stagger the number of frames that make up the checksum.

So one checksum takes 60 frames and the next takes less or more frames. Then if they edit it the frames have staggered checksum bits in it that can be seen and flagged. One part of one checksum in their and another part of a different checksum in their, so the ID makes use of them trying to blend in and in actuality losing their camourfalge and being clearly visable.

Depending on the genre of movie each kind of genre can have it's own stagger method so when it's searched it has other geners that create confusion for the searchers.
They will look for a ID in the stagger so they clearly see it, make it invisable and make them visable = drm wins over bad pirates. :)

1.) The Read and Write to check each sum in how many movies is really fast. Or one sum to check many movies?

2.) If many read write to one filebase for how many movies is too slow? Then maybe get one sum from site and check it in many movies using one read write.

3.) No idea what your asking? Is you asking how to guess the read write for matching checksum? Perhaps the database online has many read write to find right checksum, and for x number of uploaders multiply the read write to find chechsum, or use one checksum and finding right checksum problem is solved.

4.) Perhaps. It depends on how much they want to use mpeg7 to see if it's crackable or not first.

5.) The response time similar to ping. If the signal is edited for a x number of time then it matches file uploading, or if it is edited to a legal period it matches legal file upload. The theif fileshare will maximize their line and so file time will be extended meaning signal of edited file will be long.

6.) Copying drm is not easy if you can't find it. Criminals find before desembling or marking as theirs. Perhaps it is not easy to find.

Ok so other then for sites like youtube this is completely and utterly useless. Hell even on youtube it's use is questionable.

Here's just a quick list of reasons why this will never be adopted:

1. It will require the movie industry to maintain a database of the original files amounting to probably the high exabyte range worth of data already and yottabytes within the next 5-10 years assuming they go ever more to HD standards. Odds are this alone would cost them several times more then they save from preventing the uploads and that's before you factor in the net connection required to make the database accessible in a timely manor.

2. ISP's complain about filesharing as is. How do you think they are gonna react when every video submission sites has to start sending every video uploaded to an external service to be compared or downloading terabytes of videos a day to compare them on the local servers. Not to mention what this will actually do to the performance of their sites.

3. How does it determine what file to compare it to on upload? Persumable the file name and server location are embedded in the file too. Small problem with that, all you need is an editing software that strips that out which means it would never get compared in the first place.

4. You can get around this whole mess by converting the original video to some other format which doesn't have the security before uploading.

5. How does it determine if it's an illegal edit vs a legal edit. There are lots of legal edits done by every day people which it probably would flag as bad making people not want to use it.

6. Assuming you're just doing editing which just changes what's seen or heard, ie editing which doesn't change the length of the file, it would probably be easy enough to make a software which reads the original file and copies the "DRM" info to the edited file.

I'm sorry but this is never gonna work without convincing the entire recording industry, all the software makers, all the website companies and all the isps that it's worth the extra overhead when it's clearly not. Personally I wouldn't be surprised if it's dropped before the standard is finalized but that said even if they leave it in it will never be mandatory, it will be an optional checkbox just like any filter or a second encoding pass.

to the pirates when they see their heroes can't break the protection? Ho ho ho :D

It's to detect (youtube) uploads:
"a system that, with just 60 frames worth of video (about two seconds, typically) can identify copyrighted video with 96 percent accuracy and a false alarm rate of one in 200,000 -- even if it was copied from digital to analog or had captions added. This process is now part of the MPEG-7 Video signature tool, apparently"

But obviously since there is a fair use stipulated by law this would in fact be partly pointless for youtube since you are allowed to use small segments of videos, youtube would still have to check each flagged segment to see if it's fair use or someone uploading a whole show/movie (and yes there are insane idiots who upload an entire movie/tvshow in 10 minutes segments to youtube and youtube watcher who watch it that way since they for some reason think it's not illegal then, I kid you not)

Plus people will figure out how it works then defeat it then put it in handy drag&drop utilities so even youtube users can use it.

Does this mean they will only allow movies with a known signature? So pirates just make an identical copy of the original.

Or does this mean they won't allow movies with a known signature? So pirates change a single pixel in each frame or whatever.

Either way NEC is onto a winner here if they can make money from it...and everyone in Japan has 128 core CPU 1TiB RAM PCs ;-)

I know how slow my mind is and such, but this whole deal is puzzling me on how it could ever detect if some freely available video has the permission of its intellectual property owner.

So at best they could find out which videos where modified, but I don't see how that would flag them as illegal.

That if the pirates suddenly decide to stick with a proprietary, DRM-infested codec.

You guys are completely missing the point here. They are trying to sell this idea to movie business execs. Get it? It doesn't matter if it actually works in the real world.

If all it does is create a signature of each frame, and an ISP check the frames signatures against known signatures, then, all one would need to do to bypass detection is simple edit each frame.

For example, change a single pixel, even minor change, to each frames, and there you have an entire movie with different signatures. Pass undetected...

A simple re-encoding would do...

Uhm sounds like this is a simple adding an MD5 to the video header, so yeah then you can check the entire video against changes, but then what? I guess they will get MS to refuse to play altered video, and bribed lawmakers to force players to not play them if the MD5 doesn't match, so then hackers will simply recalculate the hash after they did the changes, this is all pretty old.
And if it actually works and they manage to keep the hash algorithm secret (next to impossible since players will have the code) then as Axiomatic commented, people will just skip over the crap.
But it will nicely expose who eagerly works along with this and who does not.

All this built in DRM will do is ensure that people will skip over using MPEG7. Good job guys, you're new format is already irrelevant.