User @spartanza wrote a thought-provoking post here about the possibility of using some sort of content ownership registry to deal with image plagiarism.
This might be a fruitful direction of thought, but the technical problems seem daunting.
- The registry must be able to verify the legitimate owner of the image. Otherwise, people will be able to claim vast swathes of existing images by writing a bot to scrape image-hosting sites. This makes a fully automated implementation of the registry impossible -- human intervention will be needed when ownership disputes arise.
- Image content must be well-defined. This means that in addition to URL's, the blockchain needs to store content hashes of images.
- The blockchain must be able to match an image against the registry. Which means we need some kind of fingerprinting or watermarking scheme.
Verifying an image with a particular hash has a fingerprint or watermark requires the verifier to have a copy of the image. Which means that all validating nodes now have to get a copy of all images for which the registry is used.
It seems very difficult to solve the problem against a sufficiently determined adversary. Which means perhaps we need to weaken the adversary -- for example by restricting them to using the steemit.com UI (or a third-party UI which has the same behavior).
So let's think about what you'd be able to do at the user-agent layer (i.e., as part of the steemit.com UI). You'd have the UI report fingerprint data for all images when submitting posts. When fetching images you would check the reported fingerprint data matches the image, otherwise the image is invalid.
Likewise, the registry might be done in a hybrid centralized/decentralized way by having each post voluntarily report which registry it will use, which ultimately determines who the earnings are shared with in cases of plagiarism. Posts submitted via steemit.com might all use the same "blessed" registry, but users using other UI's, or working at the blockchain level would be free to use another registry, or none at all.
We also have to have a way for an owner to deny permission to use an image -- some owners might not be willing to have their work used even if revenue is shared.
The amount of architecture that has to be built to integrate the idea of an image registry is immense, but it seems like it may be doable. It would probably have to be done in stages over many months.
It's an interesting idea, but with everything else on our plate, I'm doubtful the core developers will make any progress on this anytime soon.