Server-Based Watermarking Speeds Ramp-ups to Early-Window VOD

Petr Peterka,  CTO, Verimatrix

Petr Peterka, CTO, Verimatrix

June 7, 2011 – Ground-bound DirecTV competitors may find it’s less daunting than anticipated to launch an early-release movie service on par with the satellite provider’s new Home Premiere offering, thanks to a new approach to enabling forensic watermarking security enhancements just introduced by Verimatrix.
 
Until now, watermarking content to identify points of potential theft in the distribution system has required that set-top boxes be equipped to support the process. Verimatrix has eliminated this requirement with its server-side StreamMark platform, which can associate the content with a specific viewer by embedding watermarks on each encoded and encrypted content stream as it leaves the server.

“Having the technological capability to insert the watermark anywhere in the distribution process between the studio and the end user before it gets decoded is very powerful,” says Verimatrix CTO Petr Peterka. “It has taken an immense amount of effort to create a server-side solution that meets the studios’ requirements.”

Industry sources report MSOs and telco TV providers are pushing hard to launch the early-window home theater service, not wanting to give up any more of the initiative than necessary despite uncertainties about just what the prospects are for the costly service. So far, the four studios participating in the service, Sony, Warner Bros., Universal and Fox, have made just four movies available for VOD release 60 days after their theatrical premier, starting with Sony’s Just Go With It on April 21 and quickly followed by Warner’s Hall Pass, Universal’s The Adjustment Bureau and Fox Searchlight’s Cedar Rapids

AT $29.99 for a 48-hour viewing period, it remains to be seen what the uptake on the service will be. But DirecTV competitors are in no mood to wait to find out. Comcast and Dish Network have acknowledged they’re in discussions with the studios to get on board, and many other players are pursing the same strategy.

But while Dish has the same advantage DirecTV has when it comes to setting up new advanced protection mechanisms at a single point of distribution, Comcast and other terrestrial providers face a difficult challenge involving complicated, space-consuming technical implementations across myriad headends.

“Big operators like Comcast and Time Warner Cable have central and regional headends spread around the country with hundreds or thousands of streaming servers,” Peterka notes. “It’s non-trivial to augment those work flows with the steps required in watermarking.”

Sequential launches of the service as each VOD distribution location is ready to support the service are not an option. “The studios don’t want to be working with one-off markets,” says a cable executive, speaking on background. “It’s all in or nothing.” The focus is on developing as large a footprint as possible to accommodate scalable marketing and customer service agendas as well as to obtain a full picture of what the real business potential is.

Watermarking represents a big new learning curve for service providers. Asked to describe just how difficult the implementation process is at a large MSO, Peterka replies, “It’s a nightmare. That’s the short answer.”

Even though the FCC’s ruling last year allowed service providers to activate the selectable output control on set-tops and TVs to prevent copying of movies, the studios felt another layer of security beyond conditional access encryption was necessary to discourage theft. As Peterka notes, with the sophistication of today’s video cameras, including even Web cams, and the big-screen clarity of TV sets, it’s possible to capture a fairly pristine copy of the content with a camera right off the screen.

With an invisible watermark identifying the point where any such effort to capture and distribute content illegally took place, content owners have a means of tracking down pirates. But the studios set nearly impossible, often conflicting targets for watermarking technology to meet their requirements, Peterka notes.

“They say the watermarks must be invisible with no artifacts,” he says. “On the other hand, it must be very easy to extract the mark with consistent, reliable results, but then, by the way, it has to be very hard for pirates to detect the watermarks. This technology isn’t trivial. We worked many years to get their trust and recognition of our technology as a solution for business models like early release windows.”

Verimatrix’s studio-approved VideoMark system for embedding marks at the end points has been on the market for some time. Creating StreamMark to enable on-the-fly watermarking of encrypted content as it is streamed from VOD servers and CDN cache points with a level of invisibility and robustness sufficient to studio standards was even harder, he says, adding. “I believe we are unique in being able to do this.”

MSOs are in a big hurry to get their systems ready to support the studios’ watermarking requirements, he says. “We’re working very closely with them to help them deploy this new business model. You’ll probably be hearing announcements very shortly from several MSOs.”

As for the watermarking system implementation process, “we approach it in phases,” he says. “Initially we extend our own servers that do some of this work without disrupting the existing flow. That works for a handful of movies that they’ll initially have.

“The follow-up,” he continues, “is integrating our platform into their existing severs, whether they’re VOD or CDNs, which minimizes additional severs and additional points of failure. That’s the second phase, which we’re working on right now.”

This involves integrating the server-based StreamMark system with various encoding systems and the VOD platforms from SeaChange, Cisco, Motorola and others. “That takes longer, Peterka says. “They have to take their existing products and add our technology, and you can see it gets complicated very quickly.” Fortunately, he adds, the cable industry follows the encoding standards set by CableLabs, so that piece is easier to integrate with.

But the integration effort is not nearly as complicated as it would be if every brand and generation of digital set-top had to be upgraded to handle the placement of watermarks on content as it is decoded and viewed at the end points.

Describing how the Verimatrix watermarking system works, Peterka says there are three steps addressed by the technology: pre-processing, embedding and extraction. Pre-processing, which takes place when content is ingested, consists of an automated analysis of the video file to determine which frames are the best candidates for marking. The system then creates a set of replacement data consisting of marked frames which will be inserted during the embedding process.

Pre-processing can be done on live as well as stored content, Peterka notes. “The content can be pushed through pretty much in real time,” he says. “It can be done by a standalone server or embedded in the encoding, in which case the pre-processing can be done well upstream in the content flow.”

The embedding process occurs when the streaming server picks up the content from storage and starts streaming it to a specific end user. At that point the StreamMark system inserts the replacement frames into the stream, marking the content with invisible numbers identifying an account or transaction number that can be traced back to that user.

Not every replacement frame created in the pre-processing step is actually inserted in the stream, which allows the system to choose a sequence for placement that is unique to each user, thereby eliminating any predictability that might be used by thieves to identify where the watermarks are. Nor is it possible to read the account number with the extraction of one mark. The number can only be identified by reading a sequence of marks appearing in a segment over several minutes of viewing time.

“This is different from the way other watermarking technologies work where the marks are embedded in the physical video rather than through replacement frames,” Peterka says. “If someone does manage to figure out where the watermarks are in a given piece of content, they won’t be able to predict the location with the next piece they try to steal.” This is essential because pirates will eliminate watermarked frames if they know where they are.

This sequencing has to be done in such a way over the course of the full video so as to ensure that the user account identification is preserved no matter how badly stolen content might be corrupted once it goes through various re-encoding and reformatting processes. “Even if watermarked frames are dropped or destroyed we have multiple opportunities to recover the identification,” he says.

All of this makes the extraction process a challenge when studios run their anti-piracy programs searching for stolen content on the Web and elsewhere. When they find a stolen piece of content they have to determine what network it was stolen from and then go to the distributor to determine what watermarking process was used.

If it’s a Verimatrix process, “they send the stolen copy to us, and we run it through our extraction tool, get the number and feed it to the operator, who then identifies the customer,” Peterka says. “What’s interesting is we also need a copy of the original content to recover the mark. It’s not a blind extraction. Even though we know how we embed the marks, if we don’t have the original copy we can’t find them. This is another level of robustness, since pirates would need the original copy to have any chance of finding the watermarks.”