# Dev gives HBO free math tips to nail Game of Thrones pirate leakers

Developer Bruno Cauet has offered HBO a series of mathematical equations that could have tracked the Game of Thrones season five leaker, or even killed the leak completely. The massively popular series thought to be HBO's most profitable production was rocked over the weekend when a leaker, thought to be a translator with an …

1. There's also the issue that the screener copies can get stolen without the screener's knowledge, copied regardless of the watermarks (you should see the bootleg markets) and then sent to all and sundry.

Plus, as noted, some pirates are determined to cover their tracks and are willing to cooperate with others to cover each other's butts by collaborating on their copies to defeat watermarks. Even audio watermarking like Cinavia has been shown to be vulnerable.

1. "Plus, as noted, some pirates are determined to cover their tracks..."

Good business sense: if your sources get nicked, they won't be able to leak to you again. I'm sure the pirates could teach the big corporates a few things about customer care.

1. As a pirate I could defeat this in seconds flat. Just drop a random number of frames from the start and end of each scene. Suddenly the file doesn't match any of the records.

The only way for a system like this to work is to have people not know what you did. You could hue shift parts of the screen slightly, modify the audio or any of a number of tricks but the key is people can't know what you did. If they don't know then they have to compare multiple copies and with leaked copies, your not likely to find multiple copies.

1. >>"As a pirate I could defeat this in seconds flat. Just drop a random number of frames from the start and end of each scene"

And I can see which frames have been dropped and add them back in. Your solution doesn't work unless the hidden information is always in start and end of each scene.

1. "And I can see which frames have been dropped and add them back in. Your solution doesn't work unless the hidden information is always in start and end of each scene."

How? If your method of identification is dropped frames and I drop more frames then you can only see the total dropped frames, not the number of frames I dropped thus you cannot work out the number you dropped to identify the leak.

It's pretty obvious

1. >>"How? If your method of identification is dropped frames and I drop more frames then you can only see the total dropped frames, not the number of frames I dropped thus"

Because I have the master list of which frames I have dropped for each recipient and can add back in any you have dropped which aren't on the list. Unless you magically coincide with the same frames by happy accident (and you have to win that lottery multiple times to really obscure the signature) then I can tell the difference between which frames you have dropped and which ones I have dropped.

Now obviously if you had access to all or many different leaked copies you could do comparisons and work out which frames I had dropped from each of them and then remove all such frames from a single copy thus anonymising it, or put ones back in making it look like it's from another. But you can only do this between sources you have copies from which returns us to the situation where you have to have compromised many recipients rather than a small number or one.

Basically, if you have only compromised one recipient, what you suggest cannot work. If you have compromised two recipients then what you suggest can obscure whether your copy originated from either of those but it doesn't help you because my inability to distinguish which one out of two gives me the same information - I know that these two recipients leaked. Your technique basically only works to obscure videos between leaked parties by which point I already have the information I want.

1. "Because I have the master list of which frames I have dropped for each recipient and can add back in any you have dropped which aren't on the list. Unless you magically coincide with the same frames by happy accident (and you have to win that lottery multiple times to really obscure the signature) then I can tell the difference between which frames you have dropped and which ones I have dropped."

And should I re-encode to a different frame rate, all is lost.

My point is that once someone knows how you encode something, they can mess it up. They key is to do a range of things and not tell people what you did. That way the only way they can tell is by comparing two leaked copies looking for differences. The odds of getting two leaked copies before the release would be almost impossible.

If done this way the leaker would be terrified that they missed a fingerprint which would incriminate them.

1. >>"And should I re-encode to a different frame rate, all is lost."

No, because your new encode will still have length variations in scenes that relate to the source copies. You can average scene lengths but that brings us back to having successfully narrowed down a short-list of those recipients that were leaks. Think of your re-encode as adding 2 to every number in a sequence - it does nothing to conceal the original pattern. To do that, you need to know which numbers in the sequence are different to other sequences and change those parts in a way that is special. And you can only do that with ones that have leaked so once again - the distributor knows which parties contributed.

>>"My point is that once someone knows how you encode something, they can mess it up."

That may be the point you are trying to make but what you keep doing is posting what you think is an easy way around this which turns out not to be. Everything you say is exactly what someone who is intelligent but lacks experience in the subject matter comes out with. The problem is that each time you do this, you assume you are right without having tested it against things in practice or against counter-points.

2. >>There's also the issue that the screener copies can get stolen without the screener's knowledge,

Just because a solution isn't perfect, doesn't mean it isn't good. Whether it is stolen from a particular recipient or they handed it over knowingly, it still narrows down your investigation a huge amount.

>>"Plus, as noted, some pirates are determined to cover their tracks and are willing to cooperate with others to cover each other's butts by collaborating on their copies to defeat watermarks"

Doesn't matter. It changes the requirement from needing one compromised source to several in order to pull off this "co-operation". You are supposing many sources to be compromised and conspiring. If there are few or only one, you have again narrowed your investigation enormously.

1. "Doesn't matter. It changes the requirement from needing one compromised source to several in order to pull off this "co-operation". You are supposing many sources to be compromised and conspiring. If there are few or only one, you have again narrowed your investigation enormously."

Thing is, if nations can cooperate on matters of mutual benefit, a group of ragtag pirates with a common goal should be able to put their heads together easily, especially since each additional leaked source (and as these and other popular series prove, their very popularity makes them hot targets for theft, especially from the inside) means one more set of tracks on the same road, messing each other up. It's much like with computer entropy: barring a super-human intelligence able to deduce a complete entropy trail and negate it, any new source you throw into the mix will usually help the cause and at worst do nothing to help or hurt it.

1. It's not a matter of whether there are a group of pirates from different regions willing and able to co-operate. That's one requirement (and not a small one given these pirates compete with each other to be the first to release stuff, btw), but it's not the only requirement. It's also a matter of having to compromise multiple sources and about the distributor knowing which ones are compromised.

Right now the studio knows only that at least one of its recipients were compromised. With this, they can say: "studio X and Y in Poland were compromised, also studio Z in the UK". They can then proceed on that basis - this is valuable information. And if it were just one recipient that were compromised they would not merely have a shortlist of suspects, they would KNOW which one it leaked from. Also, it is harder to compromise many recipients than one.

You seemed determined to argue against this on a basis of lack of perfection. In fact, this is a very good and useful solution and the possible ways you point out to defeat this are partial and also more and more difficult the more recipients you hope to compromise.

2. Just because a solution isn't perfect, doesn't mean it isn't good.

Indeed, since this is a security issue, there is no perfect solution. All the defenders can do is increase the cost (typically by increasing the work factor, but forcing them to share profits - which includes credit - also works) for the attackers (pirates). If a defense asymmetrically favors the defenders - that is, it increases cost for the attackers significantly more than it does for the defenders - then it's worth using.

But, as usual, most of the Reg commentators refuse to acknowledge the most basic principles of security, such as threat models and relative costs, in favor of making banal, sophomoric claims about why someone else's idea is stupid.

1. >>"But, as usual, most of the Reg commentators refuse to acknowledge the most basic principles of security, such as threat models and relative costs, in favor of making banal, sophomoric claims about why someone else's idea is stupid."

But if someone else is stupid, it means you must be the pointer out must be smart... right?

Right?

3. #### There's also the issue that the screener copies can get stolen without the screener's knowledge

Happened in a couple of recent Oscars didn't it??

I seem to remember the 'source' of one leak was the person presenting the Oscars.

2. #### Watermarking assembly

(Dragging this from the depths of my memory so apologies if it is misremembered.)

There used to be a shareware x86 assembler that claimed in the docs that it watermarked your output. It said that there were multiple x86 instructions that performed the same operation, so by choosing them in a predictable pattern during assembly, disassembling the output could determine if you were licensed or not.

a86

3. #### I've always thought of:

Just modify a couple of pixels (not together) in a couple of frames slightly, each unique to the release.

Not sure how compression would affect it though.

A bit like the yellow dots on laser printers.

1. #### Re: I've always thought of:

It takes 1/20 of a second to print page on a printer, how long would it take to add markers to video, pack it and write to DVD ?

1. #### Re: I've always thought of:

The point is that MPEG video compression is lossy, so watermarks either have to play by MPEG's rules or risk being degraded beyond usability. That said, some screeners are willing to use destructive artifacts such as a burned "THIS IS A SCREENER" subtitle pasted periodically in the video. I suppose it depends on how far the producer is willing to go to detect or defeat screener pirates, since customizing each encode for each screener means you have to encode the movie multiple times, depending on how sophisticated your tools are (at the least, each altered section needs to be re-encoded and grafted onto the original stream).

1. #### Re: I've always thought of:

The point is that MPEG video compression is lossy, so watermarks either have to play by MPEG's rules or risk being degraded beyond usability.

There are better ways than tiny changes - large but subtle changes will pass through even heavy compression unhindered. Consider four different versions of the famous "Lena" test image I've prepared here (Safe for work, the naughty bits are cropped out). You can look at any of them in isolation and they appear quite natural. It is only when they are compared closely side by side that you can see the brighness curves of each have been subtly altered. Assuming that the image isn't compressed to the point that it would be unwatchable the differences between them will still be readily discernible. Apply that kind of filtering consistently to all the frames in a given shot and you would never know it has been done, but equally it is very difficult to get rid of without manually applying a different set of manipulations to each and every shot.

I would think something like that is a far better options than randomly inserting and deleting frames which sounds simple but I suspect would cause problems in the general case with the audio - maintaining lip sync wouln't be a problem between scenes but the musical score often extends between scenes and has to still match up very definitely with the on screen action.

Something like James Bond is the classic example for that sort of situation - the score may start gently as he makes his getaway on e.g. skis. When the bad guys start shooting at him the music responds instantly. It does so again when he skis over a cliff edge. There's a final flourish just as the parachute opens. If you do anything that alters the timing of the on screen action you ruin the dramatic effect for the reviewer, or you create problems later for the team dubbing it into Foreign.

4. #### Ummmm

Errrrr. This has been used in printed documents for at least 300 years, yah de yah de yah

5. #### multiple sources

so if the leakers have multiple sources and multiplex their release from these on a five minute timeslot basis we get a completely new "watermark" worthless to the copyright holders.

1. #### Re: multiple sources

If the leakers have multiple copies then HBO's security is really buggered. Given that I've not seen many leaks for GoT before, I'd be surprised if there's more than a couple of people leaking.

1. #### Re: multiple sources

"If the leakers have multiple copies then HBO's security is really buggered."

Or the leakers are just that damned determined, much as the paparazzi were when the Twelfth Doctor was being planned and the BBC realized they couldn't hide the news for long.

2. #### Re: multiple sources

Even worst if they could interlace the frames.

6. Thought it was deliberately leaked as an advert.

1. 4 episodes is a bit much for that, if it had been the first episode or first and second at low quality maybe, but 4 is a decent chunk of the series and wouldn't really get you much more publicity than 1

1. #### "4 episodes is a bit much for that..."

I would have thought that four of ten is precisely optimum to build the addiction. Who would dedicate time to watch four episodes, and then not be curious about the next six? Four is precisely ideal for an intentional leak marketing ploy.

Do these four episodes end with a cliff-hanger mystery?

1. #### Re: "4 episodes is a bit much for that..."

Most GoT episodes end with you asking what happens next

Episode 1 finished at a point where you ask what happens next.

Episode 4 finished with you saying with those people dead what's going to happen to her.

Answer (I think) Fly out on the back of her only free Dragon

2. Sure, the new season was going so unnoticed in the press that they needed this.

7. Surely they could just make sure each copy sent out had a different total number of nipples in it?

Would this be defeated by the pirates randomly adding, say, 0 to 3 frames to the end of each scene?

1. #### Re: Randomly add frames

My thoughts exactly. Although in order to defeat it, the pirate would have to know that the method of watermarking is a frame count. Maybe it's some coloured dots. Maybe it's a bit of audio 2 hours in that says "THIS COPY BELONGS TO FRED" in the audio. Who is going to watch 2 hours of this rubbish in order to be able to find that. I think the point is that it's easy to defeat an obvious visual watermark.

More importantly they were stupid to let such a large quantity of a valuable asset out in one chunk to a single 3rd party. Maybe next time they will employ more than one translation company and only give them half of each episode each (or something). Or get them to work in-house if that's too much of a risk. Or give them a really bad quality copy at 320x160 or something. Either way the error wasn't in the watermarking it was in trusting a 3rd party with the crown jewels in the first place.

1. #### Re: Randomly add frames

Doesn't work very well for translations, believe me. You need consistency. Imagine that you have 3 episodes, sent to 3 translators. The first translates "constructor" as "carpenter", the 2nd as "brick layer" and the 3rd as "welder". All of a sudden in the 3rd episode, you find out the killer was the welder and you are left wondering who the he'll that is when you haven't heard of them before. For the same reason, you need a few episodes after the current one to work out how to translate things that evolve in the plot.

I often watch films with audio and subtitles in different languages, and some films lose all meaning because of botched translations.

1. #### Re: Randomly add frames

How about using some reversed audio for a particular sound effect?

eg. ID number spoken, reversed and blended into a sound effect for a wheel-cart going over a bump?

or what about a blacksmith hammering out an id number in morse code?

There are a number of opportunities available.

2. #### Re: Randomly add frames

I often watch films with audio and subtitles in different languages, and some films lose all meaning because of botched translations.

My brother has a bootleg (I assume) of the HK cult classic Xian Si Jue, aka Duel to the Death, which is both dubbed and subtitled in English - with different scripts. The scripts both more or less follow the plot of the original, but the dialog is often different and some characters have different names.

Makes for an entertainingly jarring experience.

Of course the film itself is pretty wildly over the top, even by period-fantasy wuxia standards.

2. #### Re: Randomly add frames

"they were stupid to let such a large quantity of a valuable asset out in one chunk to a single 3rd party."

Strangely enough, given the total cost of this shite, it was probably down to... cost. They chose a one-stop shop that promised them heaven for rock-bottom price... so they got it.

2. #### Re: Randomly add frames

They could delete frames, perhaps. To add frames, they'd have to know what was in them.

9. "I think that binge-watching the first four episodes is a stupid idea that will make you ache for a month waiting for the fifth episode"

surely that should be, "I think that watching this dross is a stupid idea, it will rot your brain and leave you obnoxious, stupid and lazy"

10. "Cauet has advice for Game of Thrones fans too: "I think that binge-watching the first four episodes is a stupid idea that will make you ache for a month waiting for the fifth episode""

Yeh, are you guys crazy? Torrent them each week like a sane person.

I somehow doubt that for fans, that have waited ten months already between seasons four and five, that a one month "ache" will hurt that much. They will probably ease the pain by watching the four episodes they have every until episode five is released.

Fans will probably only need to wait another week for the next six episodes to leak anyway...

1. The other six episodes are sitting on the ftp server waiting for release.

2. I binged all 4 episodes and thought it was a bit weak, slow and the body count was woefully lacking.

4 week wait for ep 5 -piece of pish

11. #### Petyr Baelish copy protection

They could put a unique watermark of his accent for every DVD sent out, it changes so much that nobody would notice.

12. #### Physical Copies

All of the other issues aside, are they still releasing physical copies for review/translation etc?

1. #### Re: Physical Copies

IIRC they're in high-def and some translators have shoddy Internet access, so it's physical or bust. Besides, even for an Internet copy, a determined foe would use an HDCP stripper combined with an HDMI recorder.

13. #### Make it big and loud

If DVD/BluRay's were mailed to individual translators, why did each custom copy not have the name of the translator scrolling across the screen in giant marquee letters, that would make each individual copy unsuitable to be pirated. And if it leaked, this would allow better traceability. Also if your name was scrolling across the copy that was sent to you, you would think a bit harder about how to protect it.

Adding subtle digital artifacts, is not always the best solution.

I would see the leaks as a failure of one or more people involved in the post production. Maybe pay them more, it is not like the actors can easily leak copies. Maybe the industry needs to implement just in time post-production.

1. #### Re: Make it big and loud

Given the time it would take to encode each one for each screener/translator, not to mention the problem that this would also make them unsuitable for pressing (and you can only get a ROM-Mark with a pressed BD), how do you make a short-run screener unsuitable for pirating?

I suspect that ANY screener/translator copy is worth pirating. I see bootlegs with burned "THIS IS A SCREENER" subtitles here and there. If pirates are willing to take blatantly-obvious watermarked copies, few things will be taboo for them.

14. #### oy!

stop having them clever ideas RIGHT NOW! :)

15. How about a massive semi transparent watermark diagonally across the whole thing?

1. They'll take it anyway. They take copies that emblazon "THIS IS A SCREENER," for crying out loud.

1. Fair point but I was thinking more "Screener: translation eng-> Turkish, translatorId: 667527"

It means a bit more overhead as you can't send out one copy to all of the translators but it makes it easier to trace and you can't blur out a water mark that covers 70% of the screen.

Cost benefit needs to be done does the extra processing to put a massive watermark cost less than the potential loss of the leak? If yes then its worth it.

1. Answer's probably no for two reasons. Making custom encodes for each of the screeners and translators is going to take a good deal of time, even with professional hardware. Second, custom encodes make the videos unsuitable for sending pressed BDs, which are the only way you can send ones protected by ROM-Marks (it's part of the spec). At least a short run of about 100 copies can be justified going through the process of making the press master.

2. " They take copies that emblazon "THIS IS A SCREENER," for crying out loud."

Yep, and copies where people are getting up in front of you to use the bog.

16. #### Obviously they forgot to include the Interpol warning

'Cause that's highly effective.

1. This post has been deleted by its author

17. #### Translation

Some simple options for getting the movie translated:

1. Get the translators to come to your studio to do the translation and don't let them in or out with any media or recording equipment. I believe Apple had this approach when previewing the Watch to some devs.

2. Remove all the bits where nobody is speaking before sending it for translation. It would probably make the movie unwatchable.

3. Obscure a significant portion of the image with a big black rectangle. Again, it would make it unwatchable.

4. Send each scene to a different translation bureau - chances of them all being dishonest is smaller.

1. #### Re: Translation

"3. Obscure a significant portion of the image with a big black rectangle."

Dubbing actors have been know to complain because they are send copies in which they can only see the mouth of the screen actors.

2. #### Re: Translation

And here are why your ideas won't work:

"1. Get the translators to come to your studio to do the translation and don't let them in or out with any media or recording equipment. I believe Apple had this approach when previewing the Watch to some devs."

They'll refuse to put down the travel expenses because it wouldn't be worth it for them. It's MUCH easier and less expensive to send a disc or hard drive than a translation team. If they're THAT paranoid, they can courier the copy with an agent from THEIR studio, with all the expenses that implies.

"2. Remove all the bits where nobody is speaking before sending it for translation. It would probably make the movie unwatchable."

Don't forget signs and other visual translations, at which point it would probably become barely watchable and worth a pirate's time.

"3. Obscure a significant portion of the image with a big black rectangle. Again, it would make it unwatchable."

It also removes key context needed for some translating to make sense. Recall that English isn't exactly the most precise of languages.

"4. Send each scene to a different translation bureau - chances of them all being dishonest is smaller."

As another poster noted, consistency is essential for a good translation, which means it has to be a single firm throughout the run or else inherent translation variations build up to result in misnterpretation which can occur at key plot points, ruining the experience.

18. Couldn't the screener just trim a few more milliseconds further off to remove the identifying telltale missing bits? Thus creating a new version with no formal record of who it went to? Just a thought.

1. Depends where they tinker with it. They can add in an extra frame between scenes, throughout the middle of the episode. Unless you have another version to compare it to, you'll never know where they've messed with it. They can even insert frames between credits - anywhere the screen is completely faded to black is easy. They could even duplicate a frame where there was action (or remove one) - chances are it that the slight jump or judder wouldn't be noticed, especially on a slow scene.

1. If the pirates obtained TWO copies, they could run a picture delta analysis to determine off frames and work from there: keeping edits from BOTH copies to throw off the forensic identification.

1. >>If the pirates obtained TWO copies, they could run a picture delta analysis to determine off frames and work from there: keeping edits from BOTH copies to throw off the forensic identification"

This has been covered in detail. The above is possible. But what it achieves is to tell the distributors that TWO studios have leaked. And which ones they were. Basically, you think using n sources hides which one of n was the leak. It doesn't, it provides a list of thise n studios that have been compromised.

1. "This has been covered in detail. The above is possible. But what it achieves is to tell the distributors that TWO studios have leaked. And which ones they were. Basically, you think using n sources hides which one of n was the leak. It doesn't, it provides a list of thise n studios that have been compromised."

That depends on how they're forensically identified and how one goes about removing the traces. If they're all "add a frame here and there," the pirates can default to "trim anywhere an extra frame is detected," which would basically whittle down the forensic tagging to the point the studio won't be able to tell which studio got raided. If they're all "cut a frame here and there," you do the reverse and extend with the same results. As for "a mix of cut and trim," if you mix them up, then it's going to be much harder to tell which copy/copies got nicked because you also stand the chance of coming close to colliding with the signature of a THIRD copy, raising the possibility of a false identification. Oh, speaking of third copies, if the pirates obtain a third copy, they can probably defeat the signature reliably by using a "two-out-of-three" rule, keeping the clip length that appears in two of the three copies (and in the event of a three-way-tie between cut, extend, and nothing, keep the nothing).

And then, like I said, there's the time investment required to make each copy forensically unique, since even professional hardware takes time to encode a 1080p video.

1. >>"That depends on how they're forensically identified and how one goes about removing the traces"

No it doesn't. You should really read some of the previous comments here. It doesn't matter how you remove the frames or pad them out, it only matters that you can only do this for the cases you are aware that they are different. And you can only do that between the compromised copies you have available. If you have two copies, you can see the differences between those two. You wont know about the differences between them and a third copy that you do not have. Because you do not know about those differences you cannot obfuscate them. Therefore your "merged" version clearly indicates that you had access to copies A and B but not C. Therefore the distributor knows which two studios were compromised.

>>which would basically whittle down the forensic tagging to the point the studio won't be able to tell which studio got raided

It doesn't work like that with them unable to work out "which" single studio got raided (or sold them out). They get a list of all the ones that did and by doing the exercises you are talking about all you are doing is ensuring that list is complete.

>>"Oh, speaking of third copies, if the pirates obtain a third copy, they can probably defeat the signature reliably by using a "two-out-of-three" rule, keeping the clip length that appears in two of the three copies (and in the event of a three-way-tie between cut, extend, and nothing, keep the nothing)."

Again, no. You are assuming that the copies do not have markers in common. That's not how this works. The pool of possible markers is huge (derived from the number of frames in the movie) and all copies will have markers in common with all but one other meaning only a complete compromise of all recipients allows one to complete obscure / remove all markers and all that tells the studio anyway is that all parties were compromised.

PLEASE, read the other comments first before responding with confidence that you know better how this works. You actually don't get this. And if I sound a little short with you it's because whenever something like this is touted there are a half-dozen or so people who all just assume that they all have spotted a flaw in this that the mathematicians haven't and rather than ask if it is one, they confidently make assertions about their way of beating this as if the creators had never considered it. It's frustrating.

1. "Again, no. You are assuming that the copies do not have markers in common. That's not how this works. The pool of possible markers is huge (derived from the number of frames in the movie) and all copies will have markers in common with all but one other meaning only a complete compromise of all recipients allows one to complete obscure / remove all markers and all that tells the studio anyway is that all parties were compromised."

What about the matter that making all those encodes will take time since they're HD and each forensically unique meaning they can't be shortcutted? Plus the fact that a one-off is not worth making a ROM-Marked pressed copy?

1. >>"What about the matter that making all those encodes will take time since they're HD and each forensically unique meaning they can't be shortcutted? Plus the fact that a one-off is not worth making a ROM-Marked pressed copy?"

That's up to the distributors in each case if it's worth their time or not. Which I guess would depend on how many recipients they had (ten, a hundred, two-thousand?). It wouldn't be worth pressing these discs en masse - you'd churn them out individually. My educated guesstimate for encoding, assuming you had semi-professional hardware, would be about an hour and a half to two hours per hour of actual program / movie. But if you wrote encoding software specifically with this in mind you could do multiple files in parallel introducing the custom markers to each as you went. You'd still need to burn those files to disc individually of course, but the encoding could, if you were serious about this, be done in about five hours for a hundred or so variations. Hardware would run you somewhere around the £3,000+ level, in large part enterprise / custom NAS. I could probably modify libav to actually do this and I'm almost tempted to go and give it a go. I wonder if anyone would actually be interested in a working version of this...? I wouldn't want to be the low-paid workers who sat there burning two-hundred blu-rays in a row, however. Though on the other hand - lots of time for reading The Register. ;)

But anyway, I can't answer this last question of yours - it's a value judgement of the distributor based on how much time they want / can afford to spend doing this stuff and how much of a concern tracking leaks actually is to them. These are things I don't know. I just do maths and software. ;) But I hope this was interesting at least.

19. #### individual copies? easy then

well if you're going to modify each individual copy then yes its easy , there would be many many ways available to "tag " each copy with intended recipient

I dont think the media companies bother doing that though.

You'd have thought it would be pretty easy in these days of automation. Eg just add a code to a few random frames.

1. #### Re: individual copies? easy then

Unless you BAKE the watermarks into the actual encode, anything you try will be easy to strip. And once you bake them in, you'll fall into the pit of having to encode the episode multiple times for each screener, which given they're 1080p will take a noticeable amount of time even with professional hardware, and even then the pirates have been noted to take watermarked copies and work on scrubbing them later.

20. #### Here's an Idea...

How about changing the broadcast model instead? You know, if the series is already finished with post-production, just have the entire series available for download/streaming from [input provider of choice here] immediately. Wait a short period afterwards, monitor the volume of actual retention of the entire series (all of those who watched the entire series run, and then possibly re-watched?), then release a limited run of the physical DVD/BRay based on those figures for those who like to collect such things.

Surely by providing what people actually want for a reasonable price, instead of pandering to advertisers wallet padding, would be better in the long run? And would possibly counter the torrent leaks?

Maybe?

1. #### Re: Here's an Idea...

That sounds dangerously like common sense sir.

We'll have none of that round here, thank you very much!

1. #### Re: Here's an Idea...

No, it's not common sense. That doesn't give enough time for advertisers to get the time they pay for, and recall that advertisers can usually pay more than any group of end-users can come up with, which is why many systems today are ad-based even when users are willing to pay (because the amount they'd have to pay to make up the difference would make them balk).

1. #### Re: Here's an Idea...

"which is why many systems today are ad-based even when users are willing to pay (because the amount they'd have to pay to make up the difference would make them balk)."

Indeed - see Sky for details.

21. Watermarking like this is a much better idea than DRM..

With this kind of system, you can allow DRM free downloads, but track down the source of any breach...

Well in theory anyway, assuming the digital watermarking is done in a way that cannot be hidden easily...

1. But watermarking of this nature is basically a variant of steganography, and there are already various techniques in the know to mangle stego. Even the vaunted Cinavia audio watermark has been shown to be vulnerable.

22. This post has been deleted by its author

23. #### LaTeX

If you want to see the equations, try the online LaTeX previewer here. Much easier to read when it's rendered, though I tried and have no idea what he's trying to say.

1. #### Re: LaTeX

I pasted it into the "Try a live demo" preview at MathJax, and it rendered quite nicely. (Was easier than firing up actual LaTeX.) It does require that you allow Javascript, though.

And no, I just skimmed over the equations, didn't bother trying to understand them. He starts off by saying "suppose k is congruent to 2 (mod 1)", which is either outside my mathematical experience or a typo for "1 (mod 2)". In either case, it wasn't a good footing to start off on. It looks like he's considering all sets of a certain size, containing integers less than a certain value, where the sum of the elements of each set is another certain value; that makes sense, given his discussion. But from there I'm not going to figure it out without spending more time than I care to.

24. #### Obvious idea.

I had this idea years ago - only my variation was using slight edits to scenes. One version has a banana in the fruit bowl, one an orange. Static changes in shots with no moving camera so they are almost trivial to apply. Sixteen of those gives you 65,536 different versions.

1. #### Re: Obvious idea.

It's a technique used for product placement these days, but IIRC they don't do it yet for HD broadcasts due to the re-encoding load this places.

25. Seems like an awful lot of trouble you went to to create those equasions.

How about spending a lot less trouble and creating equasions that result in widely and fairly distributed content rather than noosing anyone who doesn't comply with your barbaric demands? Then you don't have to worry about who pirates what.

No? Don't like the sensible approach? Fine, you can continue watching everyone pirate it then.

1. >>"How about spending a lot less trouble and creating equasions that result in widely and fairly distributed content rather than noosing anyone who doesn't comply with your barbaric demands?"

I had a go at such an equation and came up with this:

Cost of Production + Profit = Fee per copy * Number of Pirates.

But I still can't come up with a value for Number of Pirates that makes it balance. Maybe if I set a negative value for Profit...?

26. #### Transcode resistant?

" audio that can resist transcoding"

I'm intrigued by this - anyone got any further sources on how audio can resist transcoding?

1. #### Re: Transcode resistant?

Probably a semantic flub. They mean audio watermarking that's resistant to transcoding like Cinavia. Most audio watermarking works on the extrema of the audio clip to avoid it being audible. However, this renders it vulnerable to mangling as I call it through simple audio transformations. Cinavia's willing to place its data in the audible part of the frequency range, resulting in a slight but barely-noticeable noise in the track. Thing is, since it's in the audible range, it's much trickier to remove without distorting the actual audio too much.

27. #### And now for something completely different !

Set up a passcode system where viewing the watermarked file requires entry of a unique code known only to the preview holder. Log the viewer's passcode entry event on a remote, well secured database (not Sony's).

If there is no internet connection available, send an OTP to the viewer's mobile phone. If there is no cell phone coverage or mobile phone, tell holders they must wait for snail mail or call a special number. My guess is they will find a cell tower PDQ.

This could provide a simple, easily enforced. tough-to-beat audit trail. However, it would also require a foolproof self-destruct mode. If the code is incorrectly entered three times (or more, depending on how intoxicated the preview owners tend to be). Also enclose a stenographic method that ensure any tampering or removal immediately destroys the file's contents. That last feature requires more analysis but there should be a way.

Of course, use of the correct codes followed by server validation would still allow at least one screen scrape (which is the fundamental flaw of any audio/video copy protection). But that still becomes prima facie evidence of insider knowledge and diffusion. Particularly if you want to be a real b*rd and force a passcode change after each viewing and then communicate it to the OTP device holder. Providing more bread crumbs to follow...peck peck.

Now, to really close the circle, make sure the preview DVD is required to phone home a set number of times before automatic self destruct. Or send the self destruct code after official film release. LIkewise, you could optionally release the DVD from self-destruct mode if the holder continues to use the generated OTPs and/or always phones home with the DVD.

When the DVD disappears from the auditing network and shows up on bit-torrent, you can audit the DB and find out which DVDs have either

a) not been officially returned to the MAFIAAA mother ship

b) not phoned home since the film's appearance on bit torrent

c) not been self-destructed

d) not been released from self destruct mode

That number should be fairly small, unless pre-view leaking is truly rampant.

It would narrow down the list of suspects in any case, and might discourage leaking, assuming that is what Hollywood really wants.

But, to be honest, I don't even know why I would want to help the film industry.

I'm still waiting for my shipment of champagne, hookers and hot-tubs. And every time I reach out, they say "don't call us, we'll call you".

Coked up gits, all of them. That's why the kind of security systems used in banking, mil-sec. many web pages, etc. will never fly there. You have to be reasonably coherent to use them.

1. #### Re: And now for something completely different !

"Set up a passcode system where viewing the watermarked file requires entry of a unique code known only to the preview holder. Log the viewer's passcode entry event on a remote, well secured database (not Sony's)."

Impossible. The disc has to be viewable at some point, and there's where ANY multimedia copy protection falls flat: pirates simply find a way to record the video during actual playback. No amount of DRM will be able to totally beat it because, at some point, it must be human-viewable for it to be useable; the pirates simply record past that point in the chain. Plus, like you said, some devices may be completely offline with way to get online, meaning there's no real chain of trust possible since an offline check against a read-only disc can be circumvented by a replay attack.

28. #### How about different watermarks?

horizontal text for screener A

vertical text for screener B

diagonal for C

removed frames for D

Hue shift for E

when a screener is leaked you just look where they bluired, no need to know what they blured, Then again I'm a pirate why Am I giving them ideas?

1. #### Re: How about different watermarks?

Multiple watermarks means multiple encodes, increasing the production time for the screeners AND reducing the viability of pressing. Furthermore, pirates can obtain multiple copies to mix and match.

29. #### If they really are serious

About stopping leaks and piracy release the damn thing when its done rather than waiting.

If the English language version was ready why sit on it?

Also regional delays are unecessary. If the latest episode of X is already out in Y then Johnny Pirate will download it.

With on demand TV im sure most people would forgo piracy and just watch it on demand. Unfortunately the choice isnt there.

Wanna see the latest Blacklist in the UK?

1. Wait for weeks or months (Or a year for the boxset).

Or

2. Pirate it now.

The improvements we've seen in everything for the last few decades have been to improve convenience. The microwave, the pop tart and so on. Movies and TV show execs dont seem to live in the same world as other people.

I rest my case.

1. #### Re: If they really are serious

>>"About stopping leaks and piracy release the damn thing when its done rather than waiting. If the English language version was ready why sit on it?"

Well it only leaked ahead of the official release by about a day so they were hardly "sitting on it". And if you're suggesting releasing all episodes at once, that's far worse from the point of view of advertising revenues so the producers would make far less money. It's also arguably worse from the public experience as for many the What Will Happen Next community factor of people getting excited waiting for the next episode, discussing it, is a big part of the experience. Millions of people discuss Game of Thrones (and enjoy doing so) in a way that simply would not happen if it were released as a big blob like a movie with breaks.

>>"Also regional delays are unecessary. If the latest episode of X is already out in Y then Johnny Pirate will download it. With on demand TV im sure most people would forgo piracy and just watch it on demand. Unfortunately the choice isnt there."

Getting rid of regional segregation and just having a single sales model for shows would certainly be a cost saving for we in the affluent West. Basic economics is you charge what the market will bare. And that figure is different in India to what it is in the USA or the UK for example. So if there's no regional segregation the price will average. That means cheaper for we in the West, but much more expensive for people in India, Pakistan or wherever.

30. Surely you'd just hire a translator to work for you in your studio?

That way you're not sending out multi-million (your currency here) DVDs around the world.

## POST COMMENT House rules

Not a member of The Register? Create a new account here.

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2020