Maybe another way to look at it.....is that the information is not in the individual pixels.......it's within the whole image. It's a distinct image quality.

Edit: The characteristic would look like random noise alone. But add a spacial reference....and you can see "in front of the scene".

 

If there aren't now, there will be at some point./--/ Every tool and technique that we devise to detect forgeries becomes the grounds for new tools and techniques to devise better forgeries.

Few would disagree, but stating so does not address the original question:

I'm not interested in the political aspects, per se, but I am interested in learning if there are examples of "faked" videos that are beyond detection.

It is not a trivial question because it, essentially, asks where we are currently at with faked videos.

When people started to make coins, they gave little thought to the edges. At some point, people noticed a practice of shaving a bit of the coin's edge. Do that with enough coins and you have some "free" precious metal. After shaving, each coin looks unaltered (barring precise manufacturing standards and precise weighing). When that practice was established, or feared to have been established, someone decided that we better make coins with a reeded edge. Shaving would remove the reeded edge and expose the alteration.

When reeded edges started showing up on coins, we can assume that, historically, shaving edges was already in practice.

If we implement a [potentially] successful form of video authentication (e.g. https://hackernoon.com/detecting-fake-video-needs-to-start-with-video-authentication-224a988996ce:

1. Hashing and signing technology to be integrated via an SDK into a recording app or onto the firmware of a recording device. It is imperative to fingerprint the video as close as possible to the time of recording as any delays increases risks of exploitation. (In the future, we may need fingerprinting built into the on-board video encoder chip or the image sensor itself.)
2. The generated fingerprints along with details such as author, location, time, and equipment details would be stored in a secure, signed and immutable manner. An example could be with a robust blockchain where a smart contract logs the hashes and the additional details of the video.
3. On playback, videos are rehashed and compared to the fingerprints retrieved from the immutable store.

We could reasonably assume that the days of faked video, passed off as genuine, are upon us.

Your point, I think, address the difficulty, if not impossibility, of creating that absolutely immutable store.

(edited for grammar and clarity)

 

As usual the devil in in the details. Almost every engineer thinks that building a secure, signed and immutable system is easy or we can use a blockchain or some other method of the day to provide authentication, security and trust in the source. This immutable store for individual image transactions at each viewing will be the bottleneck and a security weak point for all transactions. It's much better to have secure individual proof of authenticity with each item as it travels and is transformed in various way so we don't need to archive hashes, fingerprints or signatures. To expect the immutable store fingerprints to match the actual displayed image requires a reduction in the best possible image transformation security during each non-faking modification during transmission and storage. For secure individual proof of authenticity we only need to publish the authenticated public keys of trusted image makers in a secure read-only manner for a much more manageable locally cached index of images to producer public keys.