The advent of 4K services over the Internet is creating new opportunities for piracy through content redistribution with a growing consensus that the gate can only be closed with the help of forensic watermarking.
Major studios have certainly come to that conclusion, having mandated watermarking to protect UHD content as set out in the specification for Next Generation Video and Enhanced Content Protection published in April 2014 by their jointly owned R&D company MovieLabs.
The studios had realized that traditional CA designed to prevent theft of service via encryption offered no protection against redistribution of content that may have been legitimately acquired, but is now being illegally monetized, either by charging or selling advertising against it.
That MovieLabs’ decision has concentrated attention on the premium movie dimension, given the potential of UHD to deliver pictures good enough to generate high quality just by camcording. After all UHD brings TV pictures up to the quality of Digital Cinema, which has proved quite good enough for pirating with a camcorder.
With premium movies there is at least a little time to act and close in on illicit sources during the lucrative early release windows. There is no such luxury for live sports broadcasting, where the value plummets almost within minutes of the broadcast, which provides an acute challenge for rights holders, broadcasters and distributors. Unless action can be taken almost immediately to shut down a pirate stream the damage has been done as far as that particular event is concerned. While there is some value in catching the perpetrator after the broadcast is over in so far that there is then one less potential pirate next time, this is really just fiddling while Rome burns. Unless there is some way of identifying pirated streams and shutting them down within minutes, then the game could be up for OTT premium sports in the 4K era.
The immediate challenge for premium sports then is not to identify the pirate, but to locate the streams that are being illegally redistributed and close them down within minutes, to deny a good experience to the people receiving them. This is where watermarking comes in by providing a means of tagging and so subsequently identifying each content stream, in principle at every step of the distribution chain. But operators and content owners should be wary of the hype coming from the leading players in watermarking, notably Civolution, Verimatrix and to an extent Kudelski’s Nagra, for the technology does little on its own. It will have to be supported by all players across the ecosystem, including content owners, operators, CDN (Content Distribution Network) providers and device makers, to work effectively.
This is likely to happen, but getting there will involve solving a variety of technical, logistical and business challenges. Watermarking advocates are fond of saying that the technical challenges have been solved, leaving just the business issues of persuading the various partners to play ball, but that is not strictly true. It is true that the challenge of inserting watermarks such that they have no impact on visual quality and are highly resistant to deletion or obfuscation, has largely been solved. In the early days of watermarking around 10 years ago it was regarded as a bit of a joke since the marks were easy to detect and also to erase from a video copy simply by comparing one that had been watermarked against another copy and subtracting the differences.
But forensic watermarking has become much harder to break, with the codes sometimes distributed across multiple frames and in such a way that it is only possible to detect the inserted codes through comparison with an unmarked source version of the file in a computationally intensive process. Several algorithms are typically used to generate and insert the code during transit, for example to record the identity of a device receiving a stream.
It is worth pointing out that although watermarking is undoubtedly very secure and robust now, it has yet to be tested in the field for live streams simply because it has not been deployed in that capacity. Past experience suggests that watermarking will be hacked, just like every other content security technology and so its success will depend on how the system responds to those hacks. The broadcasting industry may well emulate recent developments in the IT field, where the trend in combating attacks of various kinds, including denial of service and viruses, has been to accept that penetration will occur and focus on rapid detection followed by appropriate remedial action, which is rather how immunity works in biology.
In the case of watermarking the remedial action could be to insert new codes into a stream the moment that attempts to compromise the system have been detected downstream. This would be the equivalent of the Royal Mint immediately recalling all banknotes and issuing new ones with different watermarks as soon as a forger had successfully printed convincing counterfeit notes incorporating the old marks.
Another key challenges in tackling live content redistribution lies in finding the content quickly enough to take down pirated streams within the required few minutes. This has nothing to do with the watermark itself and everything to do with the surrounding ecosystem, for detecting that a stream has been repurposed for retransmission. This is one area where technology comes in, since such action will depend firstly on algorithms that can accurately detect and then locate sources of such repurposing, presumably by seeking signals of unusual activity such as leaving a device tuned permanently into a single premium sports channel. Secondly the infrastructure has to support this process fast enough to bring back the results within a few minutes.
Then there is the question of where and how to deploy watermarking such that streams, devices and users can be identified accurately and quickly. Watermarks can be inserted at the server end or on client devices, or both, and by different players in the chain such as CDNs and ISPs. In the case of on demand content, the situation is simpler because content is delivered end to end on a unicast basis, so that watermarks only need to be session based, inserted just at the server end, which is easier to manage and more secure.
But for live sports IP multicast is commonly deployed up stream to minimize network bandwidth consumption and in that case watermarks need to be inserted at the point where the stream breaks out of the network for delivery one to one to the end user. This means watermarks either have to be inserted at the device, or at the effective multicast edge server, which may be in the domain of a local broadband provider or a CDN. If the watermark is inserted in the device this still may not offer total protection, because the user will often legitimately be able to access a sports channel say via several devices at once. The user may be able to toggle between them, unless robust procedures are in place to identify such activity. Ideally it should be possible to identify the user, not just the device.
Another challenge, common to all forms of security, is in finding the right balance between protection and convenience. Significantly, at least one of the studios behind MovieLabs, Sony Pictures, has stated that the balance must be tilted towards convenience and that some leakage must be tolerated in order to ensure continuous good service to legitimate online consumers. This applies equally to on demand and live content, with the risk being that overzealous response to an identified illegal redistribution might involve shutting down altogether at source the steam that is being repurposed, at least until the offender has been located and targeted directly. The Sony Pictures argument is that this should only be done when an alternative stream for legitimate users has been set up.
Another point that is often overlooked is the cost of licensing the technology, which was once an issue for video encoding as well as encryption. Many key patents for audio and video watermarking are held by Digimarc, based in Oregon, US, although many of these were handed over for marketing to patents collector Intellectual Ventures in October 2010 in a $36 million deal. Verimatrix originally licensed Digimarc technology although has since developed intellectual property in the area itself for its VideoMark watermarking technology, which is quite widely deployed for IPTV and more recently OTT services.
But at present watermarking technology is expensive to deploy at any scale, whoever supplies the system, largely because so much of the technology is rooted in Digimarc patents. For watermarking to be widely deployed there will have to be a licensing scheme comparable to that which evolved when MPEG-4 emerged over a decade ago. The key move came in 2002 with the joint announcement from relevant patent holders that the per player fee for MPEG-4 would be just $0.25, 10% of the $2.50 that been levied for MPEG-2, although with a new use fee of $0.02 per hour charged to video service providers.
But despite these issues it looks like the major sports broadcasters will embrace forensic watermarking, as can be seen by scouring ads for post-production staff from the likes of NBC Universal and Sky.