5 ms·

This reminds me of a crypto application I've been waiting for someone to build / explain why it would suck: Secure enclave based verifiable recording devices.

This reminds me of a crypto application I've been waiting for someone to build / explain why it would suck: Secure enclave based verifiable recording devices. So, a camera that uploads a signed hash of every image it takes using mobile connectivity as soon as it takes it. Or the same for video and audio recordings. It's loosely inspired by Town Crier[0]. The notion is that instead of regarding arbitrary blocks of data as "recordings" we demote those to arbitrary blocks of data, and anything you want to claim as verified truth needs a time-stamped hash proving what device took it, where it was taken, when, and that it hasn't since been modified. Disclaimer: I'm the furthest thing from a crypto person, this just seems like a nice use case to me. [0]: http://www.town-crier.org/ http://www.town-crier.org/

rspeer 9y ago
Well, it wouldn't help much with the topic of this article. Maybe cryptographically verifying that a porn video happened is someone's fetish, but I don't think most people involved in the production or consumption of porn would care.
- DiThi 9y ago
  It may help e.g. if the original person has an alibi at the time the fake one was made, or to distinguish which of two version came first.
DiThi 9y ago
Sounds like a good idea, if you also receive a freshly made hash: to ensure the picture was taken _after_ the received hash and _before_ it was sent out. I would make a merkle-tree-like hash to be able to obscure parts of the picture and still be able to prove the date.
Retric 9y ago
This is about half way to a block chain. The normal risk is the service simply replaces an old hash for video with a new hash secretly. However, if you put data + old hash you can create a long verifiable chain and someone can record a specific block and know the chain is unchanged from the point they saved a specific hash.
acdha 9y ago
At least Nikon has offered a similar idea for awhile: http://imaging.nikon.com/lineup/software/img_auth/index.htm http://imaging.nikon.com/lineup/software/img_auth/index.htm The catch is that it has been hacked before: https://blog.elcomsoft.com/2011/04/nikon-image-authentication-system-compromised/ https://blog.elcomsoft.com/2011/04/nikon-image-authenticatio... The Secure Enclave helps a lot but the Nikon hsck illustrates a hazard of needing devices to be updated, especially since there’s an obvious opening for someone to use a plausible old vulnerable device – and it seems like, say, a compromising photo of a politician would spread a lot further and faster than the observation that the older phone has been compromised. One other challenge is that you might be able to prove that a device took the photo but not that, say, it was a real scene rather than a high quality print or that the time stamp wasn’t subject to a manually set clock. Again, I’d expect countermeasures but worry that nobody would see them in time for the most damaging cases.
- nickparker 9y ago
  The time stamp is handled on receipt of the hash by third parties in my mind, not by the originating device. Ideally the position component of the data would also be verified by 3rd parties, eg cell towers receiving the data add their own signatures so you know it was at least in the region it claims. The high quality prints problem is more interesting. I’d hope that sort of trick would be detectable by inspection, but then we’re back to cat/mouse games with AI. If we all switched to crypto verified light field photography, we could at least force hoaxsters to build dioramas :)
  - acdha 9y ago
    One risk: all of that could make a forgery more persuasive. Imagine some James O’Keefe style operative preparing a fake but not uploading it until they drive by the victim’s house, rent a room in the same hotel or near their office, etc. How many people would see the extra metadata and decide it proved accuracy?
breakingcups 9y ago
What prevents me from pointing this camera at a 4k screen?

rspeer 9y ago

Well, it wouldn't help much with the topic of this article. Maybe cryptographically verifying that a porn video happened is someone's fetish, but I don't think most people involved in the production or consumption of porn would care.

DiThi 9y ago
It may help e.g. if the original person has an alibi at the time the fake one was made, or to distinguish which of two version came first.

DiThi 9y ago

Sounds like a good idea, if you also receive a freshly made hash: to ensure the picture was taken _after_ the received hash and _before_ it was sent out. I would make a merkle-tree-like hash to be able to obscure parts of the picture and still be able to prove the date.

Retric 9y ago

This is about half way to a block chain. The normal risk is the service simply replaces an old hash for video with a new hash secretly. However, if you put data + old hash you can create a long verifiable chain and someone can record a specific block and know the chain is unchanged from the point they saved a specific hash.

acdha 9y ago

At least Nikon has offered a similar idea for awhile: http://imaging.nikon.com/lineup/software/img_auth/index.htm http://imaging.nikon.com/lineup/software/img_auth/index.htm The catch is that it has been hacked before: https://blog.elcomsoft.com/2011/04/nikon-image-authentication-system-compromised/ https://blog.elcomsoft.com/2011/04/nikon-image-authenticatio... The Secure Enclave helps a lot but the Nikon hsck illustrates a hazard of needing devices to be updated, especially since there’s an obvious opening for someone to use a plausible old vulnerable device – and it seems like, say, a compromising photo of a politician would spread a lot further and faster than the observation that the older phone has been compromised. One other challenge is that you might be able to prove that a device took the photo but not that, say, it was a real scene rather than a high quality print or that the time stamp wasn’t subject to a manually set clock. Again, I’d expect countermeasures but worry that nobody would see them in time for the most damaging cases.

nickparker 9y ago
The time stamp is handled on receipt of the hash by third parties in my mind, not by the originating device. Ideally the position component of the data would also be verified by 3rd parties, eg cell towers receiving the data add their own signatures so you know it was at least in the region it claims. The high quality prints problem is more interesting. I’d hope that sort of trick would be detectable by inspection, but then we’re back to cat/mouse games with AI. If we all switched to crypto verified light field photography, we could at least force hoaxsters to build dioramas :)
- acdha 9y ago
  One risk: all of that could make a forgery more persuasive. Imagine some James O’Keefe style operative preparing a fake but not uploading it until they drive by the victim’s house, rent a room in the same hotel or near their office, etc. How many people would see the extra metadata and decide it proved accuracy?

breakingcups 9y ago

What prevents me from pointing this camera at a 4k screen?