Hi Sampo, Thanks for your normal post to this list. I'm not a cryptographer but I find your idea interesting. I think these ideas came up a lot when people were imagining making cryptocurrency coins and bills. On Sun, Aug 1, 2021, 7:37 PM Sampo Syreeni <decoy@iki.fi> wrote:
Many years ago I posted an idea of mine about cryptographic physical watermarking of things, such as paper money, or maybe missiles. Whatnot.
The idea was that you'd do some chaotic physical process in order to lay down a physical watermark, then image it, and finally digitally sign what was seen via asymmmetric cryptography. I imagined you'd then print the signature "on the bill" as a 2D barcode, to be verified. But I never worked out how you would deal with the inevitable "broken bill". I thought it'd take some kind of high end error correcting code.
Now it finally came to me you don't need that at all. Instead, just repeat what you imaged from the bill, on the bill, verbatim, using whatever level of ECC you want, and then the signature. The verifier can utilize the digital, error corrected replica for hard crypto purposes, while separately verifying that it matches -- in any soft statistical knee -- a hard to mechanically replicate, unique signature, embedded in the "bill".
When I read this, I imagined a use case that wouldn't work for: when the secure features are too high definition for the printing media to fully reproduce. I think then it would make sense to make multiple signatures of different regions of the secure structures, and spread the signatures across the object.
My first and best idea about how to make this physical nonce is to mix a couple of dozen well cut differently fluorescent plastic fibers into the paper or plastic fiber pulp from which the bill/artifact is made. It would be rather difficult to replicate such a random arrangement of many fibers over, say, a 1200dpi scanned bill. Especially if the highest end scanner interferometrically made sure, that the fibers/chaff really are embedded in the paper, instead of having been printed on it.
I'm imagining studying errors made in the manufacturing process. Like the micropatterns of ink placement that rely on the relative arrangement of the underlying fibers, that we've all seen looking at printed material under a microscope. I don't imagine those things are securely random, but I imagine with study you could find secure sets of them. I like to think about errors because they get more detailed than what can be intentionally produced, and exist in all manufactured objects. Everything contains artefacts of the machine used to make it, and the age of that machine. I expect the flourescent fibers are a better idea. I've just had similar ideas. Any problems with my idea? I'd like to hear, especially since it has
been a couple of decades coming.
Sorry that I don't really know the discipline. I'm thinking a first step on implementing something like this might involve automating scanning of objects at very high resolution. A digital microscope could do two dimensions. I imagine there have been a lot of advances in getting high resolution 3d information. --
Sampo Syreeni, aka decoy - decoy@iki.fi, http://decoy.iki.fi/front +358-40-3751464 <http://decoy.iki.fi/front+358-40-3751464>, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2