Verimatrix Labs: A Brief History of Forensic Watermarking in the 21st Century
At the recent Society of Motion Picture and Television Engineers (SMPTE) Annual Technical Conference and Exhibition in Los Angeles, I was invited to summarize the development of digital watermarking over the past 15 years, which apart from showing my age, is a good time to reflect on the development. Although the advancements in digital watermarking initially developed slowly, the technology is now building momentum as industry standards continue to evolve.
Forensic watermarking has previously been an afterthought in the design and architecture of content distribution. However, there are signs of this changing with some more standards being introduced to define watermarking requirements and processes, such as Digital Cinema, Movie Labs requirements and MPEG Sample Variants in the ISO Base Media File Format. With these, watermarking technology is beginning to play a more important role in video revenue security than ever before.
The progression of watermarking technology initially faced challenges due to its robustness. Many have questioned how content can be imperceptibly modified to protect against transformation that not only strips meta information, but also removes less visible information (compression) and changes everything from color to geometry and from pixel order to frame rate (camcording). Interesting research has emerged from common use and public test which indicates that algorithms do in fact exist that satisfy today’s requirements quite well.
The next challenge has been to conduct watermarking in way that is either lean and secure enough to run in a client playback device or efficient enough to run at the head-end where marks are applied before content is delivered. The most effective approach is a secure integration within the client device where any incoming content can be marked, regardless of the delivery method. However, this approach requires the longest integration time due to the integration in system-on-chips (SoCs) which require a hardware manufacturing process.
At the head-end, integration is possible with shorter lead time. While there is no control required within the client device, any update of the codec, container or delivery method requires an update to ensure that the technology is capable of understanding the format and can modify it efficiently. This is difficult because there are many updates from AVC to HEVC, from smooth streaming to DASH, from VOD server to a CDN, and even from transport stream to ISOBMFF.
While this is ongoing work, the path has been paved, and I am looking forward to tackling the next challenge presented by watermarking, whether it be advancing security in light of new attacks, supporting new application scenarios with different information to be embedded, or finding novel ways to read and react to the information found.
More information on how to adapt to many different ways content can be distributed, encoded, wrapped, cached and delivered is presented in a paper from the event. While it was not always easy to determine the best solution, for most common scenarios, there are real world solutions that have been deployed. You can download the paper (© SMPTE*) here for more details.
What’s your view on the past, present and future of forensic watermarking? Let’s discuss in the comments below or set up a time to connect.
This is the first blog post in a new series we are calling Verimatrix Labs. The blog series, written by our Security Team, will focus on how threat scenarios are trending, how they are driving new revenue security solutions, and ultimately how next-generation video service providers can create more durable business models through highly secure and frictionless services.
*Thorwirth N., “Enabling Watermarking in a Diverse Content Distribution Infrastructure,” SMPTE 2015 Annual Technical Conference & Exhibition, Hollywood, CA, 2015. Copyright © SMPTE. www.smpte.org.