If you make your fake video after the event, the location and time will be obviously off.
It would be basically the macOS Developer ID software model applied to photos.
It boils down to trust, not whether the original video was real or not.
To put it another way, reputable news organizations already go to great lengths to ensure that their stories are well sourced. At the end of the day, they are putting their reputation on the line.
In other words, 2 main stereoscopic sensors, each recording their cryptostream to their storage. In addition to that, you have 2 UV sensors doing the same thing, and 2 IR sensors as well?
The idea being that it's much, much harder to fake something if you have to take IR and UV into account.
Isn't this also where something like a blockchain is useful? Where data is dumped into it realtime by this mythical secure camera, with all its associated metadata.
Then you can ask: "does the time the data was added agree with the time the camera says the data was recorded?" "Does the blockchain agree that the data hasn't been modified after it arrived?"
Add to that the possibility that there are multiple cameras all recording this incredibly detailed video from different angles, and the evil counterfeiters will (hopefully) have a dickens of a time faking all this.
Although with all this wonderful processing, networking and storage power to enable this secure camera, you can probably create a faking system too.
Philosophically though, I think it's a good thing that many are learning not to trust information from self-proclaimed authorities :-)
If there are no known exploits, a signed video, marked as having a network based timestamp, is at a minimum incrementally more reliable than a video without such metadata.
What might be worth trying would be some sort of verified badge if this became common enough that players could meaningfully highlight its absence, similar to the way browsers are now showing non-HTTPS sites as insecure. If Facebook badged the video with a big “this iPhone video doesn’t have the usual signature” message that might slow people down.
For the deepfake porn or fake news problem I also wonder whether there’s a useful fuzzy ID service where a player could display a “this is substantially similar to <source video> and may be a copy” warning which wouldn’t be so inaccurate that nobody would heed it. That could be really important if we started seeing political fakes where someone is shown accurately but their words are faked.
Facebook / Twitter / etc verify the crypto signatures on media ingress. Then they can give proper attribution and show a checkmark or something that attests the media hasn't been altered.
So in your feed, there will be some media with checkmarks and some without and you can choose which to believe more.
Alternatively, a competitor could come along and reject all media that doesn't pass cryptographic verification.
Of course, depending on how that's implemented it could just be another vulnerability in easy-to-use proofing.
But if there was some way to add an abstraction on top of the raw keys, keeping them still accessible if the user is keen, then that might be a step in the right direction.
The otherwise is already covered by DSLRs. Canon has it, and I'm sure Nikon does. Law enforcement has been using this stuff for years.
http://www.canon.co.jp/imaging/osk/osk-e3/index.html
Apparently it was broken in 2010 (found while googling for this).
Apple signing key -> your phone's key -> signature over hash of the media (edit: for video, embed signatures every so often to you can verify a shorter clip out of the whole). Embed Apple's signature and your pubkey in the metadata of the image, and TwitYouFace will verify on upload that it's authentic.
Of course, you can't fix stupid unless you talk about a complete overhaul of public education... And even then, people are people.