It’ll probably be stored in something like a TPM, whose primary purpose is to make intact extraction of the keys difficult or impossible. A few keys might become compromised but in this scenario (unlike DRM decryption) it’s easy to ignore those keys. There’s always the chance an exploit becomes available and is more widely used, though, in which case it would definitely be less valuable.
I had many of the same questions. I have not investigated further. I’m sure some enterprising hacker will figure out how to hack it like they do everything else.
What prevents the AI from putting a hash in the photo?
Does it get validated online so that the camera company keeps a copy of the hash on their end? (Which is also problematic.)
The hash would (hopefully) be authenticated. If you want to google it, search for “HMAC”.
This is assuming that the local key doesn’t get leaked, which is assuming a lot.
Aren’t these hardware keys unusable outside the hardware?
You’d need to somehow have the AI authenticate the image through the cameras hardware to use it.
Still possible though.
It’ll probably be stored in something like a TPM, whose primary purpose is to make intact extraction of the keys difficult or impossible. A few keys might become compromised but in this scenario (unlike DRM decryption) it’s easy to ignore those keys. There’s always the chance an exploit becomes available and is more widely used, though, in which case it would definitely be less valuable.
I had many of the same questions. I have not investigated further. I’m sure some enterprising hacker will figure out how to hack it like they do everything else.