This is tilting at windmills. If someone has physical possession of a piece of hardware, you should assume that it's been compromised down to the silicon, no matter what clever tricks they've tried to stymie hackers with. Also, the analog hole will always exist. Just generate a deepfake and then take a picture of it.
You have it backwards. This is not too stop fake photos, despite the awful headline. It's to attempt to provide a chain of custody and attestation. "I trust tom only takes real photos, and I can see this thing came from Tom"
And if the credentials get published to a suitable public timestamped database you can also say "we know this photo existed in this form at this specific time." One of the examples mentioned in the article is the situation where that hospital got blown up in Gaza and Israel posted video of Hamas launching rockets to try to prove that Hamas did it, and the lack of a reliable timestamp on the video made it somewhat useless. If the video had been taken with something that published certificates within minutes of making it that would have settled the question.
If only I knew how to create my own firmware for Leica... then I could call the same crypto-chip and sign any picture I'd like. (Oh wait! There's a github for hacking Leica M8 firmware!)
Not at all. From what I understand of this article, it wouldn't stop you from doing anything you wanted with the image. It just generates a signed certificate at the moment the picture is taken that authenticates that that particular image existed at that particular time. You can copy the image if you like.
I think this is probably great for specific forensic work and similar but the problem with deepfakes isn't that people can't determine their veracity. The problem is that people see a picture online and don't bother to even check. We have news sources that care about being accurate and trustworthy yet people just choose to ignore them and believe what they want.
"that it’s a true representation of what someone saw."
Someone please correct me if I'm wrong but photography has never ever ever been a "true" representation of what you took a picture of.
Photography is right up there with statistics in its potential for "true" information to be used to draw misleading or false conclusions. I predict that a picture with this technology may carry along with it the authority to impose a reality that's actually not true by pointing to this built-in encryption to say "see? the picture is real" when the deception was actually carried out by the framing or timing of the picture, as has been done often throughout history.
You're talking about "the whole truth". If the whole is true, then all of the parts are true, so photographing only a subset of the truth (framing) is still true. If a series of events are true, then each event is true, so taking a picture at a certain time (timing) is also true.
Photos capture real photons that were present at real scenes and turn them into grids of pixels. Real photographs are all "true". Photoshop and AI don't need photons and can generate pixels from nothing.
As I understand it, it's a digital signature scheme where the raw image is signed at the camera, and modifications in compliant software are signed as well. So it's not so much "this picture is 100% real, no backsies". Nor is it "We know all the things done to this picture", as I doubt people who modify these photos want us to know what they are modifying.
So it's more like "This picture has been modified, like all pictures are, but we can prove how many times it was touched, and who touched it". They might even be able to prove when all that stuff happened.
Even that doesn't do much to prove the image is an authentic representation of anything. People have been staging photos for as long as there have been photos, and no camera can guard against that.
So basically I would just have to screenshot the image or export it to a new file type that doesn't support their fancy encryption and then I can do whatever I want with the photo?
The point is that they can show anybody interested the original with the signature from the camera.
The problem is that you can likely attack the camera's security chip to sign any photo, as internally the photo would come from the cmos without any signing and the camera would sign it before writing it to storage.
It's signed, not encrypted. Think of it as a chain of custody mark. The original photo was signed by person X, and then edited by news source Y. The validity of that chain can be verified, and the reliability judged based on that.
Effectively it ties the veracity and accuracy of the photo to a few given parties. E.g. a photo from a known good war photographer, edited under the "New Your Times" newspaper's licence would carry a lot more weight than a random unsigned photo found online, or one published by a random online rag print.
Maybe I am misunderstanding here, but what is going to stop anyone from just editing the photo anyway? There will still be a valid certificate attached. You can change the metadata to match the cert details. So... ??
I don't know about this specific product but in general a digital signature is generated based on the content being signed, so any change to the content will make the signature invalid. It's the whole point of using a signature.
Which means that there isn't a way to edit the photo and have the cert match, and also no way to compress or change the file encoding without invalidating the cert.
I'm not expert in encryption, but I think you could store a key in the device that encrypts the hash, then that encrypted hash is verified by Leica servers?
Well that's a suprise, a system that actually is comperable to block chain in a different medium doesn't plaster it everywhere. We've certainly seen more use over much much less relevance.
And where do you see any resemblance to a blockchain?
From the article it is just cryptographic signing - once by the camera with its built-in key and once on changes by the CAI tool which has its own key.
Informacam has a similar "chain of custody" goal but was developed for existing devices. Guardian Project was involved with CameraV, the android version for mobile devices. It looks like
Proofmode is now the active project & it's available for ios as well as android. https://proofmode.org/
It's one of the most obvious uses for it, I've suggested this sort of thing many times in threads where people demand "name one actually practical use for blockchains." Of course so many people have a fundamental hatred of all things blockchain at this point that it's probably best not to advertise it now. Just say what it can do for you and leave the details in the documentation for people to dig for if they really want to know.