As adaptive streaming becomes one of the main mechanisms for delivering premium multimedia content over the Internet, it is essential that watermarking works seamlessly in this environment to remain effective. In this technical paper we describe a solution that uniquely watermarks content for each receiving device in adaptive streaming applications.
Internet content redistribution piracy is the top threat for many content providers around the world. Many Internet multimedia applications today distribute content using adaptive streaming protocols (e.g. HTTP Live Streaming, Microsoft SmoothStreaming, MPEG-DASH), which allow a client device to intelligently select a content representation to dynamically react to varying transmission and/or processing conditions.
The above-mentioned protocols use DRM technologies to protect content from unauthorized access. A DRM system, however, does not protect content from piracy after it is decrypted, nor does it help find the individuals responsible. Forensic watermarking, a process of hiding data inside the content in a non-removable manner and linking this data to the user’s identity, is a security mechanism that makes it possible to trace the pirate who copies content after it is decrypted and illegally distribute it.
The existing watermarking solutions that are specifically designed to support adaptive streaming use the protocol’s ability to switch between alternative content representations to mix two (or more) differently watermarked copies of the content to create a user-specific watermark. The mixing of the watermarked copies is, typically, either done via customization of description file/playlist (a file that lists the available content representations, their properties, and rules for forming a request to access each segment), or by implementing a client-side agent. The former solution may not be commercially viable for a system with millions of users, whereas the latter solution may not be desirable, since it requires putting a critical security feature on devices in the public domain. There is a need to find a scalable and robust watermarking solution that will work with most of the Internet delivery protocols and does not require changes to any existing infrastructures.
The watermarking solution presented in “A Watermarking System for Adaptive Streaming” paper offers content providers a truly practical and easy-to-deploy security mechanism to defeat Internet content redistribution piracy in adaptive streaming environments. The solution offers the following key benefits.
Protocol-agnostic design to support existing clients. It supports any in-field adaptive streaming client, such as HTTP Live Streaming and Microsoft SmoothStreaming clients. It also supports multiple protocols simultaneously.
- Outstanding scalability to support Internet content consumption for both encrypted (DRM- protected) and decrypted content. It does not require content decrypting and decoding to make the watermark unique. The session manager deployed on a network device in the distribution network makes the content unique to each user. Content providers can distribute unique, pre-watermarked and (optionally) pre-encrypted content to a massive number of users over the Internet.
- Ease of deployment and management to ensure the best security. As a centralized watermarking solution, all intensive processing and management is done at the content provider side (head-end), without exposing any critical security function on client devices that may be prone to attack. It makes the solution easy to deploy and secure.
- CDN support to enable efficient worldwide video delivery. The solution takes advantage of the caching infrastructures in content providers' CDNs to maximize efficiency, and does not require excessive storage space since it uses just two content copies to create a massive number of unique versions for users. It enables content providers to easily deliver content over the Internet to anywhere in the world, and allows them to switch CDNs without any changes to the watermarking solution.
- Flexible security control to meet content providers' needs. The solution can be tuned to different levels of security through obfuscation/personalization of the media presentation description file. For example, the file can be the same for all clients and be distributed by a CDN, the session manager, or any other third- party. Alternatively, the file can be personalized (and encrypted, if desired) and distributed by the session manager or a trusted third party to each client individually for granular control.
The forensic information embedded in the watermark can contain many data types in binary form. It can contain unique information specific to an individual, or information about the playback environment to give content providers the intelligence needed to defeat content redistribution piracy.
The article is by Dmitri Jarnikov, Erik Hietbrink, Mark Arana and Jeroen M. Doumen, Irdeto, the Netherlands; Eindhoven University of Technology, the Netherlands; and The Walt Disney Studios. A more detailed version of this paper was originally published at the 2013 IEEE International Conference on Consumer Electronics.