Fingerprinting or watermarking? Is there a difference and does it make a difference? Anti-piracy specialist Richard Atkinson explains what they are and how they work but, most of all, why they are important for content management – and leveraging it for monetization – as much as content protection.
In comparing watermarking and fingerprinting, some aspects are similar and some are different. But, more importantly, as they continue to evolve, those differences are becoming less important. The main issue will be how it is used, because the business is driving things: it’s not just about content protection anymore, but about content leverage. With these capabilities, I can know exactly what the content is (whether there’s a mark in it or whether there’s metadata about it)...and leverage that knowledge in real-time.
Leveraging content means not just stopping bad people from doing bad things with it: we can take those consumers and offer them other, legal, ways to enjoy content. It is important to stay open to evolving applications while focusing on your business needs and requirements, because ultimately it’s technology that supports business requirements.
SIMILARITIES AND DIFFERENCES
If we look first at the similarities between fingerprinting and watermarking, one thing in common is what I call the four-step process. That means
1. Processing the content
2. Detecting the watermark and/or the fingerprint
3. Comparing reference data if it’s relative
4. Taking action.
But depending on whose products or brochures you’re looking at, you’ll commonly see these four steps depicted as three: I mark it, I detect it, and I take action.
Content management is evolving as these technologies become more mature, because I can start to use them embedded within my library and other asset management and workflow tools as, again, ways to get more out of my content. And the Holy Grail is, ‘How do I make money with it?’
One key difference is that fingerprinting, which is highly accurate, does not modify the content. It is used in the content to generate a reference file and from that reference file I can find copies of that content. It’s highly accurate.
Watermarking modifies the content; so once it gets watermarked it is not the same, there is something in there, whether you can see it or not, or whether you can hear it or not, while that’s important, the key is that the content changed through the process.
WATERMARKING
Let’s look at visible watermarking as in Figure 1. I put something like that in, and pirates are going to either mask it out through some kind of obscuring technique, or just crop it off. So it’s pretty clear that visible watermarks, at least the overt kind, aren’t necessarily the greatest. Ultimately the pirates cut them off, so that’s why as an industry we went to more of a covert marking technique.
This technique is something that one of the studios developed; it’s visible but, unless you knew where to look, how to look, when to look, you’re not going to see it. Disney’s Pixar also got a lot more creative in how they did things, they literally embedded things within the films that they would change depending on the type of individual watermarks they needed.
Figure 2 is a frame from the movie, WALL- E, where the number on the right side is completely contextual to the film and so it doesn’t stand out. That number itself was changed individually so that all the copies of the movie were traceable. One of the powerful things here is that you can see it: you know it’s there and you know it’s a watermark. There’s a certain confidence in being able to see something, and I’ll come back to that.
From the same movie, Figure 3 is a case of marks on top of marks, on top of marks, on top of marks. So early in the post-production process, as an industry the good news is everybody’s focused on security. The bad news is that that they’re just watermarking the crap out of it, and good luck if you’re trying to figure out what scene this is and actually get a creative experience from it. But you certainly know whose it was and that it wasn’t open for duplication, reproduction, distribution, blah, blah, blah.
With invisible watermarking, obviously if it’s invisible, you can’t see it, and that’s the good thing: it doesn’t really change the content from a creative standpoint, on the other hand the bad news is, is it in there?
Is it in there; I know I ran the machine and it’s supposed to be in there, but is it really in there? The good news is that technology is maturing to the point where those kind of questions become less and less of an issue. For instance, Verance can provide an inaudible audio watermark tied into AACS, aka the security model of Blu-ray, and this is another application that’s evolved over time. [The Verance watermark contains information that is inaudible to people but can be read and understood by consumer and computer devices. The mark will impart usage information about the content, including content identification, forensic tracking and copy control information.]
So, I put a watermark in my content; the consumer device then listens for a watermark and if it hears it, a certain kind of mark, it can just stop playing. This is the first time you can really take these technologies and put them right into a consumer experience, and you’re using it for essentially an anti-piracy approach, at least the defensive application on the left side in Figure 4. If the content is pirated, the player can stop the playback or mute the audio, but if the content is determined to be legitimate, it can just be played as normal.
That means no impact for any normal application; on the other hand if it was pirated material, you definitely get an experience. The experience can be that it just stops playing and says ‘I’m stopping playing because I’ve detected the watermark that says this is pirated’. On the other hand, you can use the very same technologies to change the rule from ‘stop playback’ to ‘Hey let’s monetize this thing’. You’ve got a consumer who obviously wants the content, wants to view your content; now it’s a question of ‘What do I want to do?’
FINGERPRINTING
With fingerprinting, some things are different, some things are the same. You take content and run it through a fingerprint generator, it doesn’t change the content, what it does generate is a fingerprint, a digital representation file, very similar to what people would call a hash. That in turn is stuck into a reference database and you do kind of the identical process on the other end.
Take a look at content sites; they can take a fingerprint generator and put it in their upload stream so all content going to the site is fingerprinted automatically and then sent to the digital fingerprint database for comparison. That can tell you whether the content is copyrighted or not and what the associated rules are, so you can block it, take it down, monetize it, things like that. If you get a site that doesn’t want to play by the rules and leaves content on their site that shouldn’t be there, you can scrape their site using the same technology and figure out what’s up there.
This can all be automated – no human in the loop eyeballing this; the technology says not only ‘That’s your film’, but ‘That two-second clip is your film and it’s this part of your film and by the way, the pirates flipped it, rotated it and maybe even tinted it, and it’s still your film’.
There are then a lot of things I can do once I’ve determined what content it is. As the technology has gotten very, very robust, it’s now less about the technology and more about the applications of the technology and how you get it integrated seamlessly into what you’re trying to do.
In terms of what you can do with it, I’m going to give you several different ways of thinking about it. On the left side of Figure 4, this is defensive: how do I eliminate the content, which means ‘fight the demand’. Automated content blocking, for example, or take down notices, or manual view and take-down.
In the middle of this, you have business insight in the middle, which is a very, very powerful application. The technology can tell you where your content is, which means that you can now have pretty good insight on what’s going on with it.
Then it’s all about ‘How do I take that demand and leverage it to make money?’ For instance, redirections: even if it’s pirated content, I can take the consumer and instead of showing him the pirate version that was uploaded, I can link him right over to the legitimate site. Or you can substitute, directly replace content – just pull all the pirate stuff down and put the legitimate trailer in its place.
We all know that the internet’s powered a lot by ads and the key with ads is, ‘Are they relative to what I want, or irrelevant?’ This puts ads right on top of the content and we can look at ad monetization. An example of that: I’m going to run ads with Universal and do a rev-share deal. It’s Universal’s content, so their content is generating them money in addition to my ad revenue. For instance, if it’s Fast Five in the window, the relevant ad placements mean you can run ads for the Fast Five DVD.
If you’ve got a consumer that clearly wants that content specifically, you can make all the the ad content on that page relevant to that. So it allows you to do a lot of different things. In Figure 5, there is a monetization example. You say, ‘Ok, somebody uploaded this Monty Python video to YouTube’, but using this technology YouTube can then figure out, oh, that’s Monty Python, it’s this particular scene. By the way, that particular show is now available on Amazon and so I’m going to run an ad right over the top of it, and if you click on that, it sends you to Amazon.com and you can buy the DVD.
CONFIDENCE IN THE TECHNOLOGY
So, in summary, it’s not a technology issue anymore: fingerprinting and watermarking are relatively mature technologies that work and have been incorporated into key industry standards like digital cinemas, DCI [Digital Cinema Initiatives] uses invisible watermark embedding. AACS for Blu-ray is embedded as part of the Verance watermark.
Now in the workflow, in terms of outsourcing, a number of the key industry players will do fingerprinting in addition to the watermarking; if you want to put it in your internal workflow and supply chain, you can do that too. A number of devices are starting to embed the ability to generate watermarking and fingerprinting as part of the transferring process. A key part of all this is metadata, which is fundamental to many of these processes.
There are some issues out there, such as content leverage by internal partners. When I was at Disney we saw that a lot in Europe: we’d send over a theatrical print in the old days with a European language dubbed on it, and then we’d see that print leveraged for home entertainment use. You wind up with content bridging particular distribution channels. If I stick a mark in that content that says ‘This is a theatrical print’ and all of a sudden it winds up in home entertainment, the very same content, with that very same mark, it starts to create some challenges.
Again, that’s something that the industry itself is starting to look at. Digital distribution and workflows solve some of that because it means getting the original master in what you need is going to be easier than the old days.
Broad distribution raises some ownership issues so as everybody starts to deal with digital distribution, and whether it’s fingerprinting or watermarking, getting through all those checks and balances sometimes gets complicated. It’s not really coming up yet today, but it’s on the horizon as potentially one of those areas. For instance, there is overmarking – if one of my key supply chain guys adds their own watermark, pretty soon I’m stepping on stuff, especially if I can’t see it.
There needs to be confidence in the technology, knowing it’s in there when it’s invisible or inaudible, that’s much less of a concern, but it’s always a nagging issue. And we need to consider an ecosystem that is party to different motivations; that’s just the reality of what we have as an industry, not all the players are equally motivated to do the same thing in the same way.
[The illustrated version of this article will appear in Issue 14 (January/February) of D2D.]
