I never understood the idea behind that. One can simply record the deepfaked content to have it "cryptographically blessed." That's what people still do to bypass DRM (just recording a screen) and it works so well that you can't even tell it's just a camera recording a screen.
Either way, that is assuming this security holds up and someone doesn't just figure out how to extract the key from the hardware. Have a look at the lengths people go to to hack into video game consoles, these camera vendors have no chance.
A single piece of hardware would take the image, then sign it. All of the chips for this would need to be integrated together as a secure element. The only way to get the image is from the hardware. Then it would be in the manufacturer's interest to keep the keys a secret. The keys could even be to a specific camera or batch of cameras. You then authenticate the images from the camera like your browser authenticates bytes from a web server. On X, the image gets posted, and then the browser can check the signature of the image against a key from the image sensor manufacturer. The browser displays "this image is authenticated by Sony" if they're the ones that made the image sensor.
Right. Sony holds the private keys; Sony* can make signed deepfakes. And nothing stops us from laundering authenticity by taking a picture of a picture.
* or, like Microsoft showed us, Sony insiders / intruders
If you're looking for a criticism, might I suggest, "won't somebody just print their photoshopped photo and then take a picture of that?". This criticism's semantic consistency would allow for such responses as "this could be remediated by signing additional data points such as time, gyroscope reading, and ...".
What I'm suggesting is about unfortunate centralization, the exact ideological antithesis of the technologies you are mock appreciating. I really wish I knew a better way. There's no distributed proof for truth.
Tying NFTs/Web3 to some future image sensor signing tech wouldn’t be an enhancement. The public sentiment about that space has shifted rapidly while scam after scam is uncovered.
Even if some of the underpinning ideas are useful, they should be reintroduced specifically for these new use cases.
The future of this tech needs to be developed in lock step with the existing web that people actually use. This means eventual uptake by major browsers, and hopefully agreement on open standards.
Comments
We need image sensors that cryptographically sign their output so that images can be attested for authenticity.
I never understood the idea behind that. One can simply record the deepfaked content to have it "cryptographically blessed." That's what people still do to bypass DRM (just recording a screen) and it works so well that you can't even tell it's just a camera recording a screen.
Either way, that is assuming this security holds up and someone doesn't just figure out how to extract the key from the hardware. Have a look at the lengths people go to to hack into video game consoles, these camera vendors have no chance.
This won’t help at all. Literally zero videos or photos you see have come directly bit for bit from a sensor.
Adobe Premiere, After Effects, and Photoshop would like a word.
Why can't I run that signature algorithm on my deepfaked video?
A single piece of hardware would take the image, then sign it. All of the chips for this would need to be integrated together as a secure element. The only way to get the image is from the hardware. Then it would be in the manufacturer's interest to keep the keys a secret. The keys could even be to a specific camera or batch of cameras. You then authenticate the images from the camera like your browser authenticates bytes from a web server. On X, the image gets posted, and then the browser can check the signature of the image against a key from the image sensor manufacturer. The browser displays "this image is authenticated by Sony" if they're the ones that made the image sensor.
Right. Sony holds the private keys; Sony* can make signed deepfakes. And nothing stops us from laundering authenticity by taking a picture of a picture.
* or, like Microsoft showed us, Sony insiders / intruders
NFT ads. Web3 here we go!
If you're looking for a criticism, might I suggest, "won't somebody just print their photoshopped photo and then take a picture of that?". This criticism's semantic consistency would allow for such responses as "this could be remediated by signing additional data points such as time, gyroscope reading, and ...".
What I'm suggesting is about unfortunate centralization, the exact ideological antithesis of the technologies you are mock appreciating. I really wish I knew a better way. There's no distributed proof for truth.
Tying NFTs/Web3 to some future image sensor signing tech wouldn’t be an enhancement. The public sentiment about that space has shifted rapidly while scam after scam is uncovered.
Even if some of the underpinning ideas are useful, they should be reintroduced specifically for these new use cases.
The future of this tech needs to be developed in lock step with the existing web that people actually use. This means eventual uptake by major browsers, and hopefully agreement on open standards.