Watch our fireside chat or read the article below
Steven Hawley
Founder & Managing Director
Piracy Monitor
Diagrams sourced through Friend MTS
Last time, we recognised video piracy as an expensive risk for video providers, and showed that once the content has arrived at its intended legitimate destination, the traditional video security techniques of Conditional Access and DRM can do nothing to stop it from being redistributed by entities that have no rights to do so.
The security shortcoming stems from the fact that only the legitimate path from origination to the point of consumption is being secured.
Credential management as an access management tool
Today’s video services are protected by a ‘front door’ that challenges the consumer to provide access credentials in the form of a user ID and a password, before being admitted to access the service.
In the days of traditional set-top boxes, before streaming services, credential sharing outside the home was relatively pointless for legitimate access. You had to be in the home, in the presence of a set-top box that was paired with the credentials, in order to gain access. But with streaming, where the consumer can be anywhere, credential abuse has become commonplace.
Password sharing and consumer video account abuse have captured the video industry’s attention in recent months and years, but, like DRM, the management of credential abuse and credential theft don’t help reduce the distribution of content once it has escaped the boundaries of a video service.
Identifying video content that has been discovered out-of-bounds
To protect the value of premium video content outside of these legitimate service boundaries, the video itself needs to be identified in a way that confirms its outermost point of legitimate use. Once that is known, infringing users and industrial-scale pirates can be identified.
To fill these gaps in protection not covered by DRM or CA, video providers can embed information into the video payload itself, which can occur at the origin, in the CDN during distribution or within the player device. Forensic watermarking has emerged as a preferred technique.
Payload information contained within the watermark can include the device IP address, session details, subscriber identifier, or other information.
While consumers can’t see the watermarks, automated analysis can. Let’s look at two watermarking methods that are common for IP streaming.
About server-side, or A/B variant watermarking
One technique, called A/B variant watermarking, is performed within the service provider’s facilities, “upstream” from the ultimate consumer at the video provider’s headend, or in the distribution network.
A/B variant replicates every streaming session into “A” and “B” streams, each of which receives a different watermark (Figure 1). These streams are then broken up into segments which are then combined into a single stream containing a unique combination of A and B segments so that no two users receive the same sequence.
Figure 1: Combining two sets of watermarked video
Source: Friend MTS. Image source: frames from (CC) Blender Foundation | mango.blender.org
Due to this dual stream approach, A/B variant watermarking is resource-intensive, and therefore costly, at the OTT headend. Each video source (every live video channel, for example) must be encoded twice and distributed simultaneously, meaning that the video provider needs two sets of encoders, and sufficient storage and origination resources to accommodate the two sets of streams.
There are certain additional security steps that are needed when implementing A/B variant watermarking. One is to ensure that the A and B segments can’t be discerned when they are received for playback. Another is protecting A/B variant watermarking from several forms of man-in-the-middle attacks. There are also challenges with how A/B variant watermarking would work in low latency live streaming situations.
In summary, A/B variant watermarking requires additional resource in the OTT headend, with associated costs, and security enhancements to increase its robustness, including against man-in-the-middle attacks.
About client-composited watermarking
An alternative to A/B variant watermarking is client-composited watermarking, where the watermarking process occurs within the consumer device. The embedded player implements a software library that is used to access a database that replies with a unique identifier. The watermark payload is converted into a pattern, similar in concept to a QR code, and then composited over the video.
Figure 2: Watermark is composited with the video frame
Source: Friend MTS. Image source: frames from (CC) Blender Foundation | mango.blender.org
The client-composited watermarking approach has multiple benefits that make it preferable to the A/B variant approach in certain situations.
One benefit is the time to detection, which can be as little as a few seconds.
In A/B variant watermarking of HLS-encoded adaptive bit-rate streams, using six-second segments, the amount of time necessary to cycle through the segments and positively identify the session could take as much as seven minutes. If segments were two seconds long, it’s still about 2 ½ minutes. This makes the A/B variant approach less effective for live sporting events where a match or a race could be over by the time the infringing user has been identified.
Another benefit is low cost.
Unlike A/B variant watermarking, there is no need to implement two sets of video processing, storage, and origination resources. Another benefit of client-composited watermarking is that the watermark generation and compositing processes use client-side software and don’t require any hardware modifications at the OTT headend.
And finally, this process works equally well with live and on-demand services as there is no added latency which, in the case with A/B variant implementations, needs to be mitigated.
So far, we’ve talked about how DRM falls short in fully protecting video content. We’ve also identified video watermarking as a way to fill these gaps, justifying client-composited watermarking as a preferred approach. In the next instalment, we’ll talk about how the source of infringing use can be identified and managed.
Check out the other articles in our series:
