While I admire Google's efforts in assisting copyright owners, I can't help thinking of the mass of false positives that are going to arise from this. Suppose I upload a video that I took of, say, my nieces dancing. That video just happens to bear a loose resemblance to a dance scene in a famous movie (say positions of the figures, background colour, lighting etc).

So this fingerprinting a) misidentifies my video as a clip from this movie; and b) either prevents me from uploading it or allows the movie company, which has NO rights to my video, to profit from it.

But of course, the major copyright pigs will be perfectly happy with this system, and since little guys like myself don't have the funds to buy our justice, it'll be too bad. And now that Google's done it, and shown it's possible, the US legal eagles won't be slow in making it law that all video hosting websites must use this same technology.

So here's waiting the day Russia, Sweden, or some other nation that still has some concept of freedom in it, hosts their own version of YouTube - and watch the world flock to it, leaving the US version to die the slow death it deserves.

Steve any fingerprinting in the modern digital age is going to be based purely on the binary codes that make up your video, they'll have sweet f*%k all to do with your content! More then likely it will involve some form video size and individual byte matching combined with a specific fingerprint database or some such. To a computer - your nieces dancing is not going to look anything like a scene from a movie unless youve somehow managed to recreate the exact colour, lighting, size, cinematics and everything else of the original movie for every single pixel for the length of the clip. Chances of that happening, about the same as Wales becoming the world's next superpower and Welsh becoming the new worldwide language! (no offence to all the welsh people out there - it just seemed a good comparison at the time!)

Your just being ridiculously alarmist with your comment!

Personally i cant see it having a huge affect. Once some hacker has a look and works out how the fingerprinting works, it will be a simple fix to remove the fingerprinting and people will go straight back to copyrighting. This is just google PRETENDING to be covering there asses...

How successful will this be against transcoding or mutations? (aka using a camcorder to create a new digital stream).

How fine grained is the fingerprinting? Every second of video? The entire frame or also translations of the picture?

I've studied some image watermarking tools. Even the more advanced algorithms could be defeated quite easily using mutations leaving the contents visually unaffected.

Fingerprinting will probably have similar faults.

I suppose if enough users become frustrated they will migrate to other sites (such as megavideo.com, a superset of youtube).

Personally I gave up being entertained by the entertainment industry, so this won't affect me.

(if reg is keeping tabs, I vote no to icons!)

How will they determine that someone who uploads a clip and says, "This is mine!" actually owns it?

Suppose I say that I own, for instance, a clip of my company's motion platform. But instead, I upload a scene from The Fifth Element.

If they don't have a human checking each submission to say, "Hey! That's not a motion simulator, it's Mila Jovovich... mmmm, Mila Jovovich..." then now I can rake in the cash whenever somebody watches that clip.

Worse yet, I could upload a clip of my competition's product and then get all of them deleted.

Or I could find a really popular video of something stupid, like a dog on a skateboard, and upload THAT (which is, even better, impossible to verify) and rake in gobs of cash.

My best guess is that these scenarios are going to force google to only 'protect' clips of 'big media' content - music videos, "Dick in a box" performances, etc, because otherwise the human intervention required would be prohibitive.

One last thing that would be pretty funny is to send them a clip of bog-standard white-on-black opening credits that change every couple of seconds. This might snag a whole bunch of stuff - then, naturally, profit. That'd be a bit easier for them to weed out via code though.

Any way you slice it, this is bound to be fun.

I don't think so - and I don't think rogue nations' disrespect for copyright is a "freedom" to be encouraged.

I'm sure Google will have teething problems, but I hope it comes right, and finally makes possible a balance between the rights of content makers and us consumers.

Good Luck to Google!

Water drowns, trees burn, and there's a plethora of things you shouldn't eat and/or be crushed by. Here in my home state of Colorado, people are routinely killed by the hazardous chemicals found in the granite boulders which have fallen from a mountainside and crushed their cars.

Bit beyond me technically but I would have thought that the easiest way of 'fingerprinting' is to use some kind of clever tagging technology like Shazam does for music <http://www.shazam.com>. As the clip's uploaded it's sampled and if it's tag fits something copyrighted then action is taken.

Maybe I should patent that....

As for the comments about home videos being mistaken for £100m films - errr I don't think so lads...

Reg icons: I think they're getting out of hand.... (although I love the coat and the Paris ones)

I can understand copyright owners are somewhat... averse to Youtube. Blocking uploads would be a pity though, as many people – like me – tend to use Youtube to sample/preview before they buy a CD/DVD, same as they do with iTunes previews. Basically that's what those low quality offerings good for in the first place. But the paying customer was never of much interest to the industry I guess.

I recall from a previous article that Google's fingerprinting technology does a motion-based analysis. This makes it resistant to reencoding and probably limits the false-positive rate, though it probably doesn't quite make it resistant to things like cropping, stretching, or mirroring the video, and it'd probably also be pretty easy to fool by simply changing the pulldown characteristics (like randomly shuffling frames around a bit, or making the left half of the video one frame ahead of the right half, or something).

That Google says it's ok to use whatever music you want to back your video up? Wasn't there a fuss about this? If I have a video in which I recreate a music video myself, with the music in the background is this fine?

This could kill YouTube?

Oh dear, how sad, never mind.

If I cared about videos of chavs beating up pensioners, skateboarders causing terrible phallic injuries to themselves, or grainy, crap sounding 5 minutes clips of feature films, I might be disappointed.

As it is, i'm more annoyed at the 20 minutes limit now imposed on Veoh content. Oh well, guess i'll just have to rent the box sets.

Or get it from the Library...

That doesn't work when stuff is reencoded. If two videos had to be the same, pixel for pixel, to be judged the same, then the system wouldn't work at all.

The only viable way to do this is by image analysis because the binary bits tell you nothing. There are multiple possible ways to do this, based on scene colouring, scene length, and adjacency of scenes. In particular, a movie will have camera cuts at a specific sequence of times which could act as a signature and a partial sequence would be identifiable. The chances of a false positive on a 13-cut segment would then be around the chances of a false positive at CDDB on a CD with 13 tracks - it happens, but you can scan 200 CDs and it will only happen about once. If you add some analysis of the content of the scenes the chances of a false positive become as low as you like, still leaving room for wide "epsilons" on what the exact average colour (say) is.

So I agree that Mr Roper was being a bit alarmist, but not at all for the reasons you give.

As for circumvention, you'd need to alter the things that create the signature. To do that you need to know the algorithm so we have a classic "security through obscurity" situation - considering the algorithm as a black box it's still perfectly feasible to get around it by trial and error. Then you codify your results as a tool and the fingerprint technology is uselessified - until they tweak it again. It's not like Google isn't used to being at war with people who try to attack their black boxes ... it's the same arms race as ever.

Remember when Google bought youTube, and many (including Reg) pundits said they were committing commercial suicide?

I was inclined to agree at the time, but after hearing about this widget, I'm wondering whether perhaps Google hadn't already thought of this when they bought youTube.

Video / Audio downloads online are fast destroying the conventional Music / Film industry, and one company who might stand a chance of monetizing online audio and video is Google through their advertising schemes. I'm sure they've been keen to get in on the action for quite some time.

This is pure genius. Google makes a living by selling advertising off the back of other people's content. Up to now, they've only been able to do with HTML. Now they can do with Music videos, TV Shows and Hollywood films. Brilliant.

Now what was all that about doing no evil again?...

When the revolution comes, you'll be the first against the wall!

"Steve any fingerprinting in the modern digital age is going to be based purely on the binary codes that make up your video, they'll have sweet f*%k all to do with your content!"

Lglethal, may I suggest you read http://www.digimarc.com/tech/docs/dmrc_mpeg4_dwm.pdf and maybe you'll understand why the content has everything to do whi what the watermark can be can how it is held in the file.

The Abstract in case you don't want to click on the link —A novel MPEG-4 compressed domain video watermarking method is proposed and its performance is studied at video bit rates ranging from 128 to 768 kb/s. The spatial spread-spectrum watermark is embedded directly to compressed

MPEG-4 bitstreams by modifying DCT coefficients. A synchronization template combats geometric attacks, such as cropping, scaling, and rotation. The method also features a gain control algorithm that adjusts the embedding strength of the watermark depending on local image characteristics, increasing watermark

robustness or, equivalently, reducing the watermark’s impact on visual quality. A drift compensator prevents the accumulation of watermark distortion and reduces watermark self-interference due to temporal prediction in inter-coded frames and AC/DC prediction in intra-coded frames. A bit-rate controller maintains the bit rate of the watermarked video within an acceptable limit. The watermark was evaluated and found to be robust against a variety of attacks, including transcoding, scaling, rotation, and noise reduction.

This is a general comment and not just directed at you Lglethal

I wish some people who leave comments on articles would at least be read up on the subject they are commenting on and not just shouting from the choir.

Sweden? Rogue nation? Mwuhahah.

Refusing to allow the subversion of natural rights, including those of privacy and safety from arbitary search and detention to satisfy an assumed "right" which is patently bestowed by no more than a legislated contract between some people and the state, does not a "rogue nation" make.

I think Google deserve some respect for having handled this issue very smartly (in business terms). They offer a service based on user provided content, a lot of which they have no right to provide.

If Google had added the ability to block videos BEFORE it added the ability to make money by NOT blocking videos it would of caused them massive problems.

Now they are hoping that rather than removing all these dance to some crap song videos, the record companies will choose to allow them and add advertising (which Google will profit from).

On top of this they managed to launch a video advertising, fingerprinting and management system less than one month late. This makes me interested to know how Google project manages because it seems to work (IT projects normally being massively late and broken).

My only real worry/concern is I know first hand just how bad Google are with dealing with mistakes that aren't pointed out by big media. Various forms of work are protected by fair use (as an example parody) and I don't have much faith Google will factor that in, and I have NO faith that if you try and report a removal of a work which was fair use they'll listen.

So Sweden and Russia are now "rogue nations"? Are they due for a regime change perhaps?

FYI: The RIAA/MPAA copyright vision is not respected anywhere in the world that is outside the US and A. And I don't see many indications of glowing respect inside either.

That is because it has been so obviously hi-jacked by the US IP cartels that only naive still don't see it.

Copyright was originally never intended to be a racketeering licence it has become.

You're correct about watermarking, but fingerprinting does not require a watermark in picture.

Forensic fingerprinting is designed to be tolerant of shape / size and bitrate changes in content and still identify.

Philips are the leaders in this kind of technology within the broadcast / film space, a high level overview of their fingerprinting solution can be found here

http://www.business-sites.philips.com/assets/Downloadablefile//Videofingerprinting_5-13507.pdf

And no, I dont work for Philips :-)

It will be interesting to see how much YouTube use drops when all the copyright-infringing material is all removed. Looking at the current top twenty or so videos on Google, at least half of them are of copyrighted material (eg sports coverage). Does Google pay for this already? Or will lots of it need to be removed.

Some content also includes shown-on-TV trailers, ads and pro-mos. Presumably these use music and so the music will have to be licenced for use in all territories, not just the one the ad was made for? Will this mean that US etc only TV ads will now dissapear?

Also, this move will encourage people who make stuff at home to start watermarking their own videos. Thus, videos that are intended to be hosetd on one site but end up on YouTube will now be spotted.

I imagine this is why Google didn't introduce this system when they took over. Imagine how much Google bad-mouthing there would have been if they suddenly removed 25% of the content (much of it really popular) the day they took over. Now they can make it look like it is the big media companies forcing their hand.

>The chances of a false positive on a 13-cut segment ....

>can scan 200 CDs and it will only happen about once.

Sure but if there's a million videos on YouTube, thats 5000 rejects.

Considering everyone has got used to the idea of "free" content from YouTube the horse has already bolted on this one. The problem is people really don't get the "free" idea: Google is raking in a fortune showing the copyrighted material of others, paying them nothing for it. You can't reconcile these two things and if someone is showing your material and making money from your work you're bound to be a bit pissed off. There is actually a strong moral argument to say you can't make money exploiting the work of other people without paying them. You can't expect copyright owners to sit back and allow this to happen, Google are just trying to pre-empt the fact YouTube can either be reformed or be slowly declawed. The whole culture of YouTube has circular dependencies; what's interesting or popular is normally copyrighted, people come to it because they don't have to pay any money and Google pays for it all through advertising revenue it doesn't have to share. Remove any one of those and the whole house of cards starts to collapse. Dishonest arguments about "freedom" just don't wash here, it is abstract nonsense as you well know as fundamentally people just don't like paying for things.

Welcome to Napster Mark 2.

Anyone who used to use the service long ago in the darken days will remember.

At least I will be able to upload my grainy, 5 minute video (with crap sound) of skateboarders de-genitalizing themselves and make some money from it!

What a wonderful world it is when I can make money from something that improves the gene pool!

Well, the current copyright system disrespects my property rights. It denies me right to use the physical object as I wish. It denies the enrichment of the public domain (and therefore steals from each and every one of us). It denies me my privacy, since HDDVD et al will want to phone home to see if I'm allowed to play this. It denies me my dignity because it assumes that I am a criminal.

There is no respect of me in copyright, so why should I respect copyrights?

You're all making mountains out of a mole hill.

There's no clever algorithmy thing involved, just a room full well trained chimpanzees scouring each and every video at x32 speed.

Why is your version of what is happening to Google correct?

Google don't police content (because they can't). Uploaders are almost entirely not innfringing copyrights. The skew is that "no copyright infringement in 1million accounts" isn't news, whereas "twenty 'pirates' on YouTube were arrested..." IS. So you self-select the worst.

Google are making most of their money off the content put there by the copyright holders. The copyright holders are OK with this. If they aren't, then they should say so. If they don't like that, then get the state to interfere by bringing criminal charges (Piracy is theft, remember, with the allusion [unstated because it's WRONG] that downloading movies is piracy), it's theft and criminal, let the police decide where they will do their work.

Oh, and if you state that you own something and don't, expect to see fraud charges returning your way. Hey, you tried to steal my content by saying that you can stop me from using it. If you don't have that right, that's fraudulent.

Oh, and Tim, in many cases the sports coverage was

a) public

b) minimal

so not infringing.

Or am I in trouble for telling you that South Africa won against Fiji?

Will this incorrecftly flag up vids with clips from copyrighted files? Under fair use you are allowed to use them for critical purposes.

If boffins can get Shazam to recognise millions of copyrighted songs from a short segment played down a mobile phone, then I reckon a similar technology could do the same with video....