To reduce the adverse effects of fake evidence (eventually called deepfake), off and on since 2004, and especially since ten years ago, I've been thinking up, NOT DETECTION methods, but prospective, pro-active methods to instrument a data stream, to increase the odds that video, audio, or other time-series data (including e.g. scientific experimental data) eventually can be proven to have been created from real life (and not faked) at the time & place claimed.
This is not just "proof of existence" by inserting hashes of content into a Merkle tree, like Haber & Stornetta implemented in 1990 and like we talked about on the Cypherpunks email list in the early 1990s, although of course that would be part of a complete system.
Nor is it relying on just the metadata that e.g. TruePic uses. I have been developing additional methods to asymptotically approach provability that the content depicted actually happened in real life. Think undetectable watermarks by multiple observers (courts want to see multiple witnesses) but it goes beyond that.
Relevant to this are efforts by CAI, C2PA, and companies like TruePic. I am hoping my methods are different enough, with little enough overlap, that we can realize them without licensing their IP. My goal is a more decentralized system that protects the privacy of witnesses.
My contact info is in my profile. I'd love to discuss this with anyone. I'd even be relieved if someone would explain that my ideas are not useful, or already have been done, or even would have unintended negative consequences for society. (It's hard to stay current in the relevant literature on signal processing, steganography, cryptography, and zero-knowledge proof, plus politics and game theory.)
I would even considering simply donating my trade secrets (such as they are) to some person or organization who impresses me as sincere and likely to move them forward, to avoid having to sit on them hands-off (again) during my tenure at (yet another) company that is about to offer me a full-time job today to work on something unrelated. While some of my ideas "toward authenticatability" may have been ahead of their time when I first wrote them up ten years ago, they probably aren't anymore, so the window is (at best) closing now.