AI, DeepFakes, and the End of Truth
Tony Aubé

It appears that most of this can be solved by suitable cryptographic algorithms built into digital video/audio-recording equipment. Imagine that a device can import a private key that is registered to someone’s specific identity (perhaps blockchain-based). Imagine that every so often the device embeds a cryptographic signature based on the given private key into the actual audio/video. Imagine that signature is embedded in a way that survives video re-coding. While difficult, I suspect that most of this is doable. In this scheme, while you can’t guarantee that a video/audio is truly authentic, you can now guarantee that it is at least unaltered in the time since having been signed by the author’s identity. If the identity is somehow trusted/known, this would alleviate the problem somewhat. What do you think?