Video watermarking technology is a key component of our award-winning Anti-Piracy Center and a vital tool in the battle against video piracy.
As video piracy has shifted away from P2P networks and illicit downloads to illegal real-time streaming, the problem of video piracy has dramatically increased. Accelerated in part by the Covid pandemic when visits to pirate-hosting websites increased by 31% in 2020 alone, recent inflationary pressures and the rising subscription costs of popular SVOD services have seen the problem become steadily more challenging.
Research firm Parks Associates estimates a projected cumulative revenue loss of $113 billion by 2027 for streaming video providers in the US alone as a result of content theft, with annual losses around the $20 billion mark.
The global increase in ultrafast broadband connections and the move towards app-based smart TVs have grown both the legitimate streaming market and the illegal one that piggybacks on it. Video piracy has gone real-time and it has become big business (incidentally, Parks also estimates that the value of fraudulent advertising delivered online to media and entertainment consumers in the US may exceed $700 million in 2027). Extrapolate all these figures out to a worldwide scale, and the potential losses to the industry are nothing short of catastrophic.
The result is that we need real-time tools to combat a real-time problem and tools that have been built from the ground up to deal with the latest trends in content piracy. Increasingly video watermarking is becoming a vital part of that picture.
What is video watermarking?
First, a quick bit of history. The first paper mill that introduced watermarking was based in Fabriano in Italy. Towards the end of the 13th century, it started to press a distinguishing pattern pressed into a sheet of wet paper — hence ‘watermark’ — to authenticate the end product. The result was thinner and thus more translucent paper in the raised areas of the mold which, when held up to the light, revealed the distinguishing pattern.
Over 700 years later, the overall goal of digital watermarking is exactly the same as developed in Fabriano: insert information that identifies where the content came from in a manner that is largely invisible in day-to-day use. The big change, especially when it comes to watermarking video content, is the move to subscriber watermarking; we want the watermark to tell us where the security of our content has been breached. Identify the source quickly enough and you can shut down the stream. As video piracy has moved to realtime streaming, especially of high-value content such as sports matches behind paywalls, so that speed has become a necessity.
Video watermarking techniques
There are three main methods of video watermarking currently in use today: Bitstream-Modification, A/B Variant, and Client-side watermarking.
Bitstream Modification effectively involves modifying selected areas of a picture in a non-impactful way (i.e. video quality is maintained) to be able to identify the viewer and session. Candidate blocks and the way they are to be manipulated are identified on the server-side and this information is sent along with the content in a metadata file. Then either client-side or CDN edge processing executes the changes to provide the unique identifying watermark. It is robust, but suffers from compute overhead, with the pre-processing stage in particular adding latency into the system, making it less useful for live content. It is also not necessarily well adapted for legacy consumer equipment.
A/B Variant watermarking is a similarly two-step process. Primarily aimed at the OTT and on-demand sectors, two video streams are created which are totally similar except for the watermark. It is how they are then interleaved — again, either client-side or via CDN edge processing — that provides the unique identifier. However, watermark extraction is slow as the identifying sequence can be long, especially for clients with large subscriber numbers, and there has been little enthusiasm in the industry for the increased delivery infrastructure costs the dual-stream approach requires.
Client-side watermarking is by far the dominant current method, and the simplest. It is favoured for its low cost implications, its rapid watermark extraction, and its deployability across multiple platforms, including legacy set-top boxes. Essentially a graphical overlay is composited onto the video stream in the client device, and that is it. This can be visible — many viewers will have seen numeric codes appearing at the edge of their pictures during high-profile premium events — or invisible, the latter being both more popular with viewers and harder for pirates to re-stream as the mark has to first be detected.
If client-side watermarking has a weakness it is that it needs to be deployed alongside a robust DRM. It is not an end-to-end solution as the watermark is not applied until it reaches the client device, therefore the content needs to be safeguarded separately up to that point (though best practises would indicate this is also the case with client-side executions). Also, as with this type of solution, care needs to be taken that the client device is tamper-proof.
What to look for from a video watermarking solution
Whatever technology is used for video watermarking — and we would tend to recommend a client-side implementation for some of the following reasons — the ideal solution will feature the following attributes:
- Dynamic adaptation
Video piracy has become an exceptionally organised and exceptionally agile activity. The time to take action and take down pirated content, especially when it comes to live sport, is while it is being streamed ― not minutes hours, or even days afterwards. That means the solution has to be able to agiley adapt to threats at the same speed as the threats iterate to find a weakness.
- Fully-deployable
Legacy devices need to be included for maximum security benefit. This approach also helps when it comes to future-proofing, meaning the solution is also deployable on future generations of devices.
- Robust
Watermarks should be retained and detectable even after a breach, and stay coupled to content during all forms of acquisition (physical recording, screen casting etc.) and via post-acquisition processes such as re-encoding, aspect ratio change, and on to storage.
- Fast
Watermark extraction needs to take place in seconds.
- Imperceptible
With picture quality an increasing cause of differentiation between operators, an ideal watermark should be imperceptible and not impinge on picture quality. It should also not increase latency, either on the server- or -client-side.
- Scalable
An effective watermarking solution needs to be able to scale to millions of users and devices with no impact on the speed of detection, especially when it comes to live sports watermarking.
Once these attributes are met, then a dynamic watermarking solution can provide an effective measure against pirate activities. It is worth pointing out though that it should not be deployed as the sole measure in any anti-piracy deployment. This is especially true given its important role in establishing where a breach has already occurred, but rather it should be seen as a component in an over-arching suite of anti-piracy tools and initiatives.
