I’m using a blockchain because it’s currently the most practical Byzantine Fault Tolerant (BFT) network available for timestamping the Initialization Vector and final hash. If you prefer a web-of-trust or another cryptographic framework, feel free to use that instead.
By projecting across the entire scene, we significantly increase the complexity of any real-time forgery attempt. A single display that barely interacts with the surroundings is comparatively easy to spoof and not particularly convenient to deploy. Of course, if you think that simpler setup is worth exploring, go ahead and experiment!
Keep in mind, this method differs from merely signing an image. It creates an optical puzzle that reality solves faster than a simulation can. The physical interaction with the environment adds complexity that’s difficult to replicate artificially. I tried placing a display in front of the camera about a decade ago, but found that approach too “analytically solvable,” so it lost its appeal for robust authentication.
Largely true, although submitting timestamp hashes to the blockchain is probably the easiest bit.
"Camera looking at a monitor": While that might be simpler in some setups, it doesn’t really solve my main issue. I want the signal to permeate the entire scene, not just appear in the corner of a display or overlaid on the video. By projecting the challenge onto all visible surfaces, we create a physical environment that’s difficult to fake (since you’d have to convincingly generate or remove those patterns in real time). Air-gapping isn’t really the goal right now.
Finally, we're need much finer granularity than 15 minutes! The point is to lower the generation time below what is achievable via generative model.
Thank you for the comment, and I hope these clarifications are useful. It's a new concept, so please forgive the clumsiness with which I may be communicating it.
Thank you! It is indeed a little like a signature based on proof-of-projection.
As they say, once you have a signature, you have a most of a cryptosystem.
I've been experimenting with those and other applications of non-linear functions in projector-camera systems.
I've developed a novel approach to creating tamper-evident video via cryptographic feedback loops between projectors and cameras. The process works as follows:
1. A projector displays a challenge pattern (Perlin noise derived from of a hash)
2. A camera captures this projection
3. The system hashes the captured image concatenated with the previous hash and uses it to derive the next projection
4. This chain demonstrates true temporal sequentiality that's difficult to forge
By incorporating random noise derived from Byzantine Fault Tolerant networks and using these networks as timestamping servers, the proofs inherit the network's decentralization properties. ML then confirms that the feature distributions in projection-photograph pairs match expected patterns from the training dataset.
A real solution is to create privacy-preserving proofs regarding the state of a timestamped cryptographic feedback loop in projector-camera systems.
It's trivial to add incoming image hashes and timestamps to a Merkle tree and to derive the output hashes of projector emissions as they come, while providing an output hash to one's interlocutor (or BFT network) periodically. This proves the beginning time, end time, and sequentiality of all emissions.
Correspondence between emissions and recordings is proven by training autoencoders on concatenated projection images and returned camera images. These autoencoders are transformed using privacy preservation techniques.
Thus, human relationships and trust can be modeled and secured.
In my opinion, yes. I posted elsewhere on this thread to spam my project, which is to derive an initialisation vector from a blockchain (or other interlocutor), derive a projection from this I.V., record an image of the resultant projection, hash it, project the hash, repeat this process, and resubmit the final hash to the blockchain. A neural network is used to verify the correspondence between projection and image (e.g. by training an autoencoder on the concatenated RBG matrices). In a mature system, this process is performed at a rate not achievable by generative networks.
A sample dataset is downloadable at truthbeam.eth over IPFS.
In its basic form the Truth Beam mechanism comprises a video projector and global shutter camera with an exposure time approximately equal to the projector's refresh rate and a computer.
An intialisation vector acquired from an interlocutor (e.g. a blockchain) is used to create a cryptographic feedback loop in this system, i.e. projector[i]=hash(camera_image[i-1],hash[i-1). The top hash of this tree is returned to the interlocutor for timestamping. This proves the approximate time of creation and sequentiality of the projector emissions.
Each cryptographic interval of this system is described by two 3-channel matrices (one for the projector image, one for the camera) and these matrices concatenated and used for training an autoencoder used for anomaly detection.
Things get more advanced later (e.g. a theoretically fully continuous system current being tested in discrete models.)
Feel free to ask any questions; I'm very interested in Hacker News' observations and I'll stick around to chat.
By projecting across the entire scene, we significantly increase the complexity of any real-time forgery attempt. A single display that barely interacts with the surroundings is comparatively easy to spoof and not particularly convenient to deploy. Of course, if you think that simpler setup is worth exploring, go ahead and experiment!
Keep in mind, this method differs from merely signing an image. It creates an optical puzzle that reality solves faster than a simulation can. The physical interaction with the environment adds complexity that’s difficult to replicate artificially. I tried placing a display in front of the camera about a decade ago, but found that approach too “analytically solvable,” so it lost its appeal for robust authentication.