Watermarking for Surveillance Tutorial

Published Aug 22, 2012 00:00 AM
PUBLIC - This article does not require an IPVM subscription. Feel free to share.

Watermarking is one of the most confusing aspects of surveillance. This can become a serious problem when watermarking is demanded by users (or courts) to prove the validity of surveillance evidence. In this note, we explain the two main ‘types' of watermarking and how watermarking is implemented in real world systems.

The Two Watermarking(s)

Unfortunately, watermarking has two common meanings that are radically different. You need to be very careful what someone means when they say ‘watermarking'.

  • Meaning 1: Image overlay - an image is placed on top of the surveillance video. The image is ‘watermarked' with an image of the company/owner/user of the surveillance system:

overlay added

  • Meaning 2: Digital hashing / digital watermarking - Every individual bit of the video is checked and a code is generated. Later, the video can be checked again. If the new code does not match the original code, the video must have been changed. This ‘watermarking' cannot be seen on the video/image but is far more sophisticated than image overlay.

Even More Advanced Watermarking?

Entertainment companies who want to protect their videos from online theft often seek even stronger watermarking protection. Instead of just proving that the video was not tampered, they want to determine who has distributed the video or where the video was distributed from. For instance, if a video is pirated, who leaked the video originally? This is sometimes called video fingerprinting. However, we have never seen a surveillance company implement this.

Choosing the Right Watermark

Since the two types are so radically different, choosing the right one for your application is easy:

  • Marketing / Branding - Use an image overlay as you want people to see that it is your organization while the risk of faking is low.
  • Evidence / Courts - Use digital hashing as you want assurance the video has not been tampered. NEVER use image overlay as a ten year old with simple software can fake that.

Where to Watermark

First, the easy and simple case - image overlay watermarking. This can be applied fairly anywhere and anytime - on most IP cameras, on most VMS systems as well as on basic video editing software even after video has been exported. The image below shows an example of this feature onboard a camera, in this case an Axis model:

overlay image

Digital Hash Watermarking Location

Where to apply a digital hash watermark is much more complex, both because of implementation constraints and evidentiary concerns.

The first question is: How early do you want (or need) the video to be watermarked? Does it have to be right away when the video is created (at the camera)? Can it wait until the video is recorded? Is it OK just to hash it when it is exported from the recorder?

Here are the options for digital hashing:

  • At the camera
  • When the recorder receives the video
  • When the recorder exports a video clip

However, in practice, statistically, digital hash watermarking is rarely done. Many, if not most courts, will accept exported surveillance video with no watermark at all. However, when digital hashes are applied it is typically applied when exporting from the recorder itself. Export configurations often have an option to enable digital watermarking.

Sometimes digital hashes can be applied when the video is first recorded but rarely is this available in IP cameras. Be especially vigilant to check any marketing claims of ‘digital watermarking' as often this only means image overlay, not digital hashes.

Digital Hash Watermarking Support in Surveillance Products

While variation exists, here are some rules of thumbs:

  • Most cameras support image overlay but very few support digital watermarking (exceptions - Mobotix and Axis cameras that use a 3rd party plugin from TRedess [link no longer available])
  • Most recorders do not support digital watermarking when the recorder receives the video (exceptions - Genetec has an option to do so but it increase CPU consumption similar to performing server side motion detection; also ipConfigure has an option using SHA512)
  • Most recorders do support digital watermarking for export though the method used can vary significantly. If you really care about digital watermarking, it may be worth checking what method they use.