Does this move us closer to needing registries of authentic images? (White-listing? Only a partial solution.)
Google's response today, in addition to removing nano banana from Earth, was that people should just use Google's synthID(?) AI detection app, to determine if this is a blacklisted (effectively) image.
Of course that's impractical because a million people could quickly receive the image .. the image would make an impression on them before detection. And, this would also mean that Google's AI detection software needs to be ran by every human for every image that they view, in order to determine if it's valid. So, let's say that's 100 images, so that's a 100 to the millionth power? (Did I really get a CS degree?) I think it's just 100 million visits to Google synthID tool.. lol.
If we go with a white listing approach then we know that we should trust the image. If we go with a blacklisting approach then we only know that we should not trust the image, which we already knew before we ran the tool....
Watermarking is for fools. Those doing the generation and watermarking will control what is perceived as false or true. Watermarks have historically been about proof of ownership or origin. Watermarking fake creations works in reverse, all of a sudden what's void of watermarking is true, but those that control the watermarking can either add them to truths to turn them into fakes or skip watermarking for certain people or organizations, making their fake creations become truths.
I've toyed with the idea of issuing LTE IMEI-like identity numbers for digital cameras (and a LTE IMSI/SIM-like identity number to identify a paying accountholder who wishes to subscribe to an image-creator attestation service).
Each photograph would attach cryptographically-verifiable metadata attesting to both the identity of the camera and the person who owns it. (Say, m="Camera 123 owned by subscriber 456 created a raw 20MP image with sha 0xdeadbeef... at time 2026-01-23 4:56pm" countersigned by both the device and subscriber keys).
Image compression would be an immediate issue. Key lifetime management would be another major issue, as would key revocation behavior. Device manufacturers would take time to buy in and phase in hardware support. Consumers would object to yet another subscription service. Privacy and freedom-of-the-press advocates would probably have something to say about establishing a norm where we doubt the legitimacy of any digital images that refuse to "out" their creator.
And other than those pesky little non-starters, I've yet to find anything wrong with it :'-)
I also think we need to consider modulating the meaning .. eg This camera is owned by XYZ person ... Since I think it could be a much more short term ownership.
Ie if I take one of these photographs and then register the media with my identity / private keys then I can assert This Is Mine.
Every media output before and after that from this camera is only circumstantially related to me.
Oh dear, I should have clarified what I meant by "identity" here.
By "attesting to ... the person who owns it", I mean to talk about "identity" in the precise sense of "identifying" the relative physical location of whatever entity happens to be in control of some particular key material issued and registered a priori at some secure location. (That is, a supercomputer can bruteforce many kinds of strong cryptography eventually... but unless P = NP, it's not happening quickly. The odds are literally astronomical that one would be able to guess the right answers correctly and quickly. So we can all rest assured that anyone who can make it through a zero-knowledge proof protocol within reasonable time limits must physically control certain private key material).
Again: I am not talking about "identity" in the usual everyday sense of identifying some particular human being.
A better way to put it would be something like "SIM cards, but for cameras". You could even use similar terminology to what LTE has:
- IMSI (International Mobile Subscriber Identity) := the smartcard from your cell carrier which lets you use a particular phone number on their network. (For the camera use-case: you would subscribe to an attestation service that will vouch for your photos. The service would issue you a phone-number-like "subscriber number", perhaps via SIM smartcard, u2f dongle, or other physical "key")
- IMEI (International Mobile Equipment Identifier) := the network-specific serial number that device manufacturers issue to each device, in addition to the regular "make-and-model" serial number that one might expect. The function of the IMEI is to label a particular physical device as end-user equipment that can participate in the attestation system. (It also comes in handy to support anti-theft blocklists in the event that user equipment is stolen, and some nefarious third party attempts to connect later).
Note that in the case of LTE devices: it is possible to clone a SIM or steal a mobile device. Fortunately, the IMSI + IMEI combo gives cellular carriers enough information to do something reasonable in those situations. For the camera use-case, an attestation service could similarly do their part: 1. Refuse service if the request IMSI is associated with a spam dialer, a closed account, or a nonpaying subscriber (whose service has been shut off); 2. Refuse service if the request IMEI is associated with a stolen device.
Again, it doesn't give you the usual notion of "identity". But (in theory) it ought to give you something similar to the notion of an in-service phone number.
Various digital cameras already support in-camera crypto-signatures created by the sensor package iirc, it’s just not the sort of thing you find in typical consumer hardware. Look into forensics specialty gear perhaps?
Comments
Does this move us closer to needing registries of authentic images? (White-listing? Only a partial solution.)
Google's response today, in addition to removing nano banana from Earth, was that people should just use Google's synthID(?) AI detection app, to determine if this is a blacklisted (effectively) image.
Of course that's impractical because a million people could quickly receive the image .. the image would make an impression on them before detection. And, this would also mean that Google's AI detection software needs to be ran by every human for every image that they view, in order to determine if it's valid. So, let's say that's 100 images, so that's a 100 to the millionth power? (Did I really get a CS degree?) I think it's just 100 million visits to Google synthID tool.. lol.
If we go with a white listing approach then we know that we should trust the image. If we go with a blacklisting approach then we only know that we should not trust the image, which we already knew before we ran the tool....
- https://www.digitaldigging.org/p/how-to-plant-a-nuclear-plan... original article is more interesting, also linked in above main article
Watermarking is for fools. Those doing the generation and watermarking will control what is perceived as false or true. Watermarks have historically been about proof of ownership or origin. Watermarking fake creations works in reverse, all of a sudden what's void of watermarking is true, but those that control the watermarking can either add them to truths to turn them into fakes or skip watermarking for certain people or organizations, making their fake creations become truths.
I've toyed with the idea of issuing LTE IMEI-like identity numbers for digital cameras (and a LTE IMSI/SIM-like identity number to identify a paying accountholder who wishes to subscribe to an image-creator attestation service).
Each photograph would attach cryptographically-verifiable metadata attesting to both the identity of the camera and the person who owns it. (Say, m="Camera 123 owned by subscriber 456 created a raw 20MP image with sha 0xdeadbeef... at time 2026-01-23 4:56pm" countersigned by both the device and subscriber keys).
Image compression would be an immediate issue. Key lifetime management would be another major issue, as would key revocation behavior. Device manufacturers would take time to buy in and phase in hardware support. Consumers would object to yet another subscription service. Privacy and freedom-of-the-press advocates would probably have something to say about establishing a norm where we doubt the legitimacy of any digital images that refuse to "out" their creator.
And other than those pesky little non-starters, I've yet to find anything wrong with it :'-)
(see also: Schneier's Law)
Very nice. I think this is important work.
I also think we need to consider modulating the meaning .. eg This camera is owned by XYZ person ... Since I think it could be a much more short term ownership.
Ie if I take one of these photographs and then register the media with my identity / private keys then I can assert This Is Mine.
Every media output before and after that from this camera is only circumstantially related to me.
Relevant:
https://en.wikipedia.org/wiki/Content_Credentials
https://en.wikipedia.org/wiki/Content_Authenticity_Initiativ...
Nice.
I see this as A stepping stone to personal digital identity.
Oh dear, I should have clarified what I meant by "identity" here.
By "attesting to ... the person who owns it", I mean to talk about "identity" in the precise sense of "identifying" the relative physical location of whatever entity happens to be in control of some particular key material issued and registered a priori at some secure location. (That is, a supercomputer can bruteforce many kinds of strong cryptography eventually... but unless P = NP, it's not happening quickly. The odds are literally astronomical that one would be able to guess the right answers correctly and quickly. So we can all rest assured that anyone who can make it through a zero-knowledge proof protocol within reasonable time limits must physically control certain private key material).
Again: I am not talking about "identity" in the usual everyday sense of identifying some particular human being.
A better way to put it would be something like "SIM cards, but for cameras". You could even use similar terminology to what LTE has:
- IMSI (International Mobile Subscriber Identity) := the smartcard from your cell carrier which lets you use a particular phone number on their network. (For the camera use-case: you would subscribe to an attestation service that will vouch for your photos. The service would issue you a phone-number-like "subscriber number", perhaps via SIM smartcard, u2f dongle, or other physical "key")
- IMEI (International Mobile Equipment Identifier) := the network-specific serial number that device manufacturers issue to each device, in addition to the regular "make-and-model" serial number that one might expect. The function of the IMEI is to label a particular physical device as end-user equipment that can participate in the attestation system. (It also comes in handy to support anti-theft blocklists in the event that user equipment is stolen, and some nefarious third party attempts to connect later).
Note that in the case of LTE devices: it is possible to clone a SIM or steal a mobile device. Fortunately, the IMSI + IMEI combo gives cellular carriers enough information to do something reasonable in those situations. For the camera use-case, an attestation service could similarly do their part: 1. Refuse service if the request IMSI is associated with a spam dialer, a closed account, or a nonpaying subscriber (whose service has been shut off); 2. Refuse service if the request IMEI is associated with a stolen device.
Again, it doesn't give you the usual notion of "identity". But (in theory) it ought to give you something similar to the notion of an in-service phone number.
Various digital cameras already support in-camera crypto-signatures created by the sensor package iirc, it’s just not the sort of thing you find in typical consumer hardware. Look into forensics specialty gear perhaps?
A bad actor takes your camera with all your crypto identity and shoots some illegal pictures or pictures of pictures.
Straight to jail.
The proof is of unmodification, not of handler; for that, you want a second sensor in the eyepiece.
Yes, exactly.