Watermarking has been around since pre-digital days. The process of embossing, through a vaguely complex method, a secondary validation of the authenticity of a fixation of content has certainly had successes. Notable is paper cash.
However, particularly in the digital world, one of the problems with watermarking is that it can be relatively simply removed. While this may be as simple as overwriting the channel ID bug with a black square, if the watermark is applied invisibly using steganography, then it can be placed anywhere in the video. To remove it, you would, broadly speaking, degrade the video quality. Depending on how the steganography is applied that may only be a small degeneration.
A bigger problem is that applying watermarks at scale is a complex task. For broadcast signals, sometimes a code has been burned in between the integrated receiver decoder (IRD) element of the receiver, and the video-rendering system. However, that pipeline can be relatively simply hacked. If you want to make your content hacker proof, then there is no point delivering him a clear signal in the first place ...
... so watermarking has traditionally struggled because all content bears the same watermark. It ultimately only proves that it was that watermarked version that was pirated:
Shameless plug here: At id3as we came up with a watermarking system that actually delivered each viewer a uniquely watermarked (through any watermark application model you want to plug in) stream at every distribution. This means that if the file is redistributed (which we think is unpreventable by force) the pirate copy can be traced back to the initial authorised user directly. If it is tampered with using other video of the same source, and if that second video is also from our network, then we will simply end up with a fingerprint of both pirates.
We think that this model places the responsibility for antipiracy response clearly at the door of the rights holders: it presents them with the data to act. Critically this in turn frees up the distribution networks from having to bear the cost of this reaction (think ISP “take down notices”) in order to fundamentally secure the business revenue strategy of a separate party - the rights holders.
Sadly disruptive technologies are not always welcome, and while I do dream our watermarking technology could create a new tactic that while it doesn't prevent piracy, it does empower rights owners, who can isolate entities who are systematically pirating and target them with direct legal action, where today you largely have to prosecute vicariously “via the ISP and Telco because of privacy law” since only the last mile networks can tie IP addresses with residencies, and they are tied up with a great deal of regulation about releasing that. Our model allows the rights owner to examine a sample of “pirated” content and directly look up who they licensed it to who's copy was allowed to be pirated without needing to go through the ISP. From a technology point of view it works for DRM free content, simplifying issues such as targeting a universal HTML5 video renderer.
Some of this will be given more context in Chapter 7.
As a final comment very few CDN architects will ultimately have to do anything with Watermarks. The rights owners take a view that the ‘DRM suffices, so lets focus on discovery' way forward is the best and has been so for some years.
Watermarking could be a disruptor in the right time, but for now the market - and in particular, the lawyers and accountants - are finding some working models with DRM, and is resistant to further major disruption.
Be ready to add a transparent “bug” if you ever have to capture video. But other than that, watermarking is, sadly, unlikely to become a widespread, and more important, a key accounting function in monetizing music or content online in the next few years. I do, however, expect it to emerge in the next 8 to 10 years as an increasingly interesting way to enforce rights - particularly of high-value premium live events.