This might be a weird analogy but this product feels like it recreates the NFT problem -- meaning that it's inherently self-referential and doesn't actually solve the problem it purports to.
NFTs are supposed to be an irrefutable, immutable record of ownership. But the only thing they can actually "prove" is the ownership of the NFT itself. It's like having a receipt that says "I own this receipt."
It's a similar story here: using a QR code is proof that at one point you got this QR code from qrdate.org, but it doesn't prove you didn't manipulate the image or video, in which case what value is it providing?
Remimds me of Vitalik Buterin holding a paper with a new block's hash and posting his photo when there were rumors about him being dead a few years ago.
> An irrefutable cryptographic timestamp is just that, irrefutable.
In math, sure, but he specifically asked about a court of law. Irrefutable in math doesn't mean admissible in court. Then there's the 1, 6, 12, or however many random people that decide the case; they likely need the math explained to them, and two sides get turns convincing them the other is wrong.
There was a case in Australia 10 or so years back, where MD5 sums were added to police speed camera photos. This was before MD5 vulnerabilities were well known.
The defence successfully argued that MD5 was not totally reliable -- even though the chance of it being faked was absolutely tiny.
The real irony was that without the MD5, the photos would probably have been accepted just on the police assertion of chain of custody.
(I've been looking for this news story but unfortunately I can't find it at the moment)
Thanks for the link- exactly the sort of thing that I was looking for in feedback. I'm aware it's conjecture, but do you think this sort of precedence would be useful in airing doubts about the trustworthiness of an Ed25519 signature in court? Do you think the wording on the site is problematic in regards to this, or would you have any thoughts on how concerns about this would be best communicated?
Most of cryptography can in theory always be brute-forced or simply guessed. There is no absolute guarantee, though you can get very very close. And there are other ways around it, such as keys getting leaked.
Also, because of lack of expertise courts may simply not want to deal with this kind of proof and thus not acknowledge it.
I believe at that point it's up to the lawyers to argue. It's up to persons with people skills to convince other people that something is not fake, using tools like these as assistance, not as clinchers.
But what would that solve? The blocks on the blockchain are public, so anyone can fake a timestamp by grabbing an old block and generating a QR code.
The point of this website is that the server signs the timestamp. It's more than just a QR code with a time and date, it has some actual verification code.
You can always "hold up an old newspaper". Nothing prevents you from storing and reusing an old signature.
It always only proves that it wasn't taken before a certain date.
You can then also prove "not after" by timestamping the picture (having it signed by a RFC3161 timestamping authority or putting a hash on the blockchain).
Good point. And the people who decide what the newspaper front page will display for a given day are constrained to some degree by newsworthy events on that day.
That thought does does cast an additional possible funny interpretation on a photo I took of the adjacent street newspaper vending boxes of the two major Boston papers one day. The Boston Globe main headline was of a serious major news event. The Boston Globe above the fold was mostly filled with a photo of some political or sports figure, and a lowbrow muckraking headline about them. A funny (highly implausible) theory is that the Herald editor/publisher had needed to determine the front page for a paper issue, days or more into the future, such as for faking proof-of-life on/after a date, and then had to run that front page. :)
> It does not work against the past (taking snapshots of the code and using them later)
Right on the page, it says this.
Maybe I am not understanding your point. Both the website and the blockchain strategy make it impossible for you to record video content and claim it happened in the future, but neither of them defend against recording after the timestamp happened.
The blockchain strategy does not require a centralized service, while the QR code requires the website to be up forever.
You can always use the published public key to verify the signature.
There's also a way to generate the codes offline using a certificate chain and a fingerprint (https://github.com/qrdate/qrdate#static-qr-date-v1-spec) which does not require a website, just a key store. It needs more specifying to be viable for real world use but it should be possible. Help is very welcome.
If you print out a piece of paper with the latest block, hold up the NYT's, or use your service, you still only prove that the content was created after a certain date.
AFAIK, signing off a timestamp and offering a public key to prove the timestamp was signed then doesn't seem to add value other than unnecessary complexity. I guess if there is a specific time you're trying to prove the event happened after, this system would offer higher fidelity than others, but not all.
This is certainly a possible implementation and good if you would want to base it on the chain, doable via a minimal node. It would not work offline though, which would make it harder to use in frontline reporting, so I think the spec itself should not specify such a requirement.
There's two proposed ways specified. The web way (called "Dynamic") and a fingerprint-based implementation that can work offline ("Static"), for example in approved apps for some organisations: https://github.com/qrdate/qrdate#static-qr-date-v1-spec
It is still a concept as there's no implementation of it besides the tests in the repo, so it would require further refining for production apps. For example with specifying key exchange formats, which we are looking at right now. Anyone with expertise in this field wanting to contribute is welcome!
Yup, I'd love the old way better. As long as it's human-readable.
I hate barcodes in a way because they require tools to read. And circumstances vary a lot!
Programatically superimposed usage of the code is problematic for all the same problems as with EXIF manipulation, so the thinking is to recommend using a physical prop (smartphone or paper) that makes it harder to fake the code in later. When combined with rapid distribution (to a news agency, live feed, Telegram or such), there's a period of time where I think it's unfeasible that someone would had tampered with it to future-date an older image.
It is possible to superimpose the code of course, as I noted:
"For example, when uploading photos through a messenger service, the service can sign the photos upon upload date and superimpose the QR Date onto it. Any further uploads to date the photo again would have to either crop the code out or otherwise mangle it so much that it would raise suspicion.
For videos, the QR Date can be superimposed as either moving around in the frame, switch places, or other motion to make it harder to superimpose another code later."
However these would only be indicative, not definite proof, and would need to be combined with other efforts- mainly deduplication, to find doubles.
For key storage and logging, I don't think it's my job to try to convince people to trust me with a private key (or logs) for the purposes of introducing this idea. Frankly I do not want the responsibility. If it ends up being useful, I would be happy to donate the domain to a reputable NGO or journalistic organisation for proper storage of the private key on a hardware device, contracts on how it and logs are handled, etc. In the meanwhile, if you want to trust the site, it's up to you just like with any CA.
I don't think a single online centralised authority would be a good idea in any case- it's why I'm trying to also consider situations where it can be used offline and by many organisations that each have their own requirements for trust.
A practical idea. I wonder whether a more consensus-based alternative to this could use some aspects of the current-highest-block from a blockchain to incorporate proof of a previously-unknowable, globally-verifiable value.
Some sort of a web of trust based on fingerprinting has been discussed, but it's an abstract idea still. Currently I don't think a blockchain would bring any added advantage- in my mind there is no need for unobserved codes that are older than a few hours unless someone emerges from a jungle or similar place with zero internet coverage to distribute the photography. The whole idea is that outside observation needs to happen rapidly for social proof to emerge, based on which journalistic organisations could make editorial decisions.
I agree that you don't need a blockchain for this but consensus systems can be nice. A system that others can run this system and somehow verify with your keys as well as their own, or anyone running this and offering the API. So scanning that QR code will send me to the website of whoever generated it but will also show other instances's approval there.
Yes, it's a good idea. It has some security and privacy considerations of its own which are outside of my area of expertise, and I believe it could be implemented on top of the existing spec.
And this could make it more useful too. You could prove that an event happened before a certain timepoint (not after it).
Some notary services allow you to inject documents or hashes into cryptocurrency blockchains which prove that the submitted documents existed before a timepoint.
> You could prove that an event happened before a certain timepoint (not after it).
Not with what GP proposes.
If I understand their proposal correctly, it is to include a value from "the current-highest-block from a blockchain to incorporate proof of a previously-unknowable, globally-verifiable value." You can always include historic values trivially.
It is much more involved to actually do a transaction on that chain, which could indeed be used to prove existence before a certain time. But yes, your alternative proposal would work for that, if you assume the chain cannot be controlled by any individual/partisan party.
> but if you include a timestamp that was signed by a trusted third party within your photo, in a reasonably non-fakeable way, it is then provable beyond a reasonable doubt that you indeed are capturing the near-present.
This smells like complete bullshit to me. You can incorporate a QR code like this into any image whatsoever.
To prove "beyond a reasonable doubt" that an image was taken at a certain date, or not before a certain date, you need a trusted witness who saw it being taken, who signs the entire image.
Yes, the observer (preferably multiple) is an integral part of the process. The second signing is an interesting idea, though I'm not sure if it would prove more veracity over, say, a newsroom verifying the code.
"verifying the code" means absolutely nothing. Any code you're verifying is from the past, and could be edited into any image from any time in the past.
All you know is that neither the image nor the code came from the future, which is self-evident.
I suspect the idea here is to rely on it being difficult to forge a printed version of the QR code that is shot together with the scene. For that purpose, a newspaper is better; it is more complex. I mean, someone could take a picture in, say, the year 2032, using a new QR code (printed on paper and everything), and then carefully edit an old QR 2022 QR code into that image. It's a very regular, high contrast, simple image.
In the era of deep fakes, I can't believe someone would think any of this is "beyond the shadow of a doubt".
A newspaper might not be available; often pictures with a blank piece of paper with the date written are used. QR Date relies on the signature in the QR code, not just the date, which makes it harder to fake within a reasonable amount of time. The proof is based on the public key being from a trusted authority, and the speed of dissemination.
I don't think anyone has said "beyond the shadow of a doubt" besides you. Reasonable doubt is a different thing. If you wanted more veracity, you could always keep on taking more pictures around the subject with more codes, or just keep holding the generator within frame for video.
The news article headlines get cut off pretty quickly to the effect of "Putin [says something]", but I guess it's not overly important as long as the headlines differ in wording.
Thanks! I thought about using <marquee> for some flair, but concluded it's best to leave it as simple as possible for now. :) The headline is there to try to provide some familiarity to the idea and it's not required for implementation.
Yeah, a marquee probably introduces the potential for blur, which can obscure the photo/video and also create false positive suspicions of manipulation if it looks blurry in the finished capture.
From UX side, nice addition would be displaying time in local time zone from browser and/or additional list to select timezone/country on page with scanned QR code
Thanks for the feedback, on the front page it was a consideration to keep it UTC to not accidentally dox anyone using it on-stream. I like the idea for having a selectable timezone on the verification page.
IMO more useful would be signing while image with verifiable timestamp. this could be used then to prove that some image is at least that old.
It could be useful when you expect that someone will try to alter it and publish it as original, you could then prove that your version is older therefore have to be original.
This is similiar to how Smart Health Codes (COVID-19 QR Code) work, they contain JWTs which link to the public key. You keep a list of valid public URLs to trust.
NFTs are supposed to be an irrefutable, immutable record of ownership. But the only thing they can actually "prove" is the ownership of the NFT itself. It's like having a receipt that says "I own this receipt."
It's a similar story here: using a QR code is proof that at one point you got this QR code from qrdate.org, but it doesn't prove you didn't manipulate the image or video, in which case what value is it providing?