What is Video Fingerprinting?
When digital content is identified without modifying the original media, it is referred to as fingerprinting.
In the same way a human fingerprint is a compact way of recognizing the identity of a single individual, a fingerprint is a compact representation of a larger media file.
For example, if a digital fingerprint is generated for the movie The Lion King, it can be used to detect copies of the same movie online or in file-sharing networks.
To detect whether a given movie is a copy of The Lion King, a fingerprint is generated for that movie too. That fingerprint can then be compared to the original fingerprint – if they are identical, the movie is a copy of The Lion King.
Ideally, a digital fingerprinting technique can measure similarity: If a copy of The Lion King is slightly modified, it will produce a fingerprint that is similar (but not identical) to the original fingerprint, by some measure of similarity.
What are the advantages and disadvantages of video fingerprinting?
The advantage of using fingerprinting in content identification, is that the original media does not need to be modified in any way. The disadvantage, however, is that a fingerprint cannot help the user distinguish between different copies of the same title.
For example, fingerprinting could be used to identify a video found online as The Lion King, but it will not be possible to determine whose copy of the movie was leaked.
Fingerprinting is used by services such as YouTube’s Content ID, which helps the service determine when copyrighted material is uploaded from accounts that do not have the rights to redistribute it. This is one major use case of fingerprinting – to efficiently find unlicensed use of media. Another use case is Automatic Content Recognition (ACR), where apps like Shazam use fingerprints of music to identify a song that is playing.
What is video watermarking?
Unlike fingerprinting, watermarking is the intentional modification of a media file to add information to it, such as the identity of a recipient. The information added to the media is called the watermark’s payload.
Further, digital watermarks can be divided into those with payloads that are human-readable (visible watermarks) and those that can only be read by computers (forensic watermarks).
Visible watermarks may be used to provide ancillary information about media (e.g. the edit version) or identify the intended recipient (e.g. to display the name of a screener reviewer). The latter is often seen in Oscar screeners, which may be marked with the receiving Academy member’s name. Forensic watermarks are used in a similar way, but are intended not to be perceptible by the human eye.
Forensic watermarking is the process of hiding identifying information in each individual copy of a video file (read our whitepaper about the topic).
A good forensic watermark is almost completely imperceptible, impossible to remove without also damaging the host video, and secure – it cannot be faked or modified. In the past, only the largest studios and distributors could afford forensic watermarking.
For media rights management, a watermark that identifies the intended recipient is most relevant. Each copy that is distributed is embedded with a unique payload that is associated with the intended recipient. This can either be their name or some identifier that is associated with them.
Visible and forensic watermarks are often both included in videos. On the one hand, embedding a visible watermark with someone’s name can be a powerful and visible indication that the person will be associated with a leak. That said, this is a blunt instrument that does not allow for a nuanced approach to dealing with rights infringers.
The same deterrent effect can be achieved by adding a visible notice that the video contains personally identifiable information.
This post is an excerpt from our latest white-paper, Why Watermarking is not Enough: How Blockchain Technology can be Used to Stop Online Piracy. Click here to download your free copy today.