This is fun. I've been designing something similar involving encryption via a custom deck of 64 cards and long term durable secret storage in base-4 as differently sized washers on a bolt (with labels so you only have bother decoding it in extreme circumstances like a house fire).
I think more cryptography should be designed with simplicity as a goal. Being unsure if you can trust the computer should not exclude one from secure communication.
Instead of a custom deck, you could use a standard 78 card tarot deck and remove or just ignore one of the 14 card suits. Now you can buy encryption decks at many game or book stores and it's something people might plausibly have sitting around the house.
Do you have any more details about the card or bolt/washer systems anywhere? They both seem like cool ideas.
The idea was that if you're going to bind significant value to an artifact that can be stolen it should be:
- kind of heavy, like... not pickpocketable
- hard to destroy, like... can survive a fire or being buried for a million years (happens in some scifi I'm writing)
- made of things you can commonly get with cash, so nobody can look at your purchase history and know you have one
The thought was that we could use it in math class. Sort of like an alternate reality where basic numeracy meant something different.
I can't quite recall the details of why, but I remember that if you do three washers per base-64-symbol, and you stick to messages which are multiples of four symbols long, then any conversions that a student is likely to to are unlikely to run into weird issues involving padding (equals signs are used by base64 converters to say "no data here, but not zero's either', I wanted to sheild the user from having to think about that).
Also, humans are good at preserving near-exact copies of prose, so if you need an entropy source you can take a particular scene in a novel and strip it of non-base64 characters, and use it (in addition to other entropy sources) as a sort of one-time-pad.
I was also working on social key recovery. It would be unreasonable in today's world to expect that average folk would be carrying out such an algorithm by hand, but supposing that it was indeed considered basic numeracy, maybe not. Especially if it's the only way to gain access to your inheritance because something dystopian has happened to the legal system. (Nowadays I'm not sure I think that generational value transfer is a good idea, but this is a snapshot of how I was thinking about it 6 years ago).
As for the cards themselves, I wanted one side to show you how to encode that base64 symbol in washers on a bolt, and the other side to show you how to add/subtract that symbol with every other base64 symbol. There were going to be encryption / decryption mats which told you where to lay the cards and how to carry out the operations.
I was going lock up some money, compromise half of the social-key-recovery scheme at a talk at defcon or somesuch, and see if the community could crack into the other half. I figured I'd just put more money in the prize wallets every year and use the "the money is still there" as a source for warm fuzzy feelings about the security of the system.
The you-cant-trust-computers-and-there-are-no-banks future (probably fiction, possibly not) that I was preparing for... it's not coming as soon as I thought at the time. I still might pick this back up and propose it for use in math classrooms and such, but I haven't been giving it much energy lately.
At some point during the design of psst, I considered using GF(4) instead of GF(5). I chose GF(5) mainly because it supports up to four shares. It's also slightly more efficient to encode a-z text, and to obtain a random digit using a dice.
I think more cryptography should be designed with simplicity as a goal. Being unsure if you can trust the computer should not exclude one from secure communication.