Unfortunately, this is still as insecure as end-to-end encryption since the author deploys it potentially on Internet-connected computers. The computer will be hacked, a keylogger installed, and plaintext recovered. It also may have the property of Perfect Forward Interception where future messages are breached via a rootkit. The necessary modification comes from high-assurance security: data diode. A one-way link. It will allow the 32-byte packets to leave the encrypting machine but no attacks can come back in. Preferably an optical one with the two at a specific distance to eliminate electrical-level attacks. With the two computers in separate, faraday cages running on batteries to cover some other attacks. :)
Two computers in separate space observatories, point the telescopes at each other and use maritime signalling.
I mean, none of this protects metadata privacy at all… with the space observatories, it’s incredibly obvious who’s talking to who. :)
Such a risk might be mitigated with the observatory equivalent of mix networks like we did with email. That means they gotta constantly look at each other plus random ones. All of them participating in even one shared message do this with a secret timing ensuring they look at the right time. Probably be out of service due to mechanical failure most of the year.
My old method of using infrared at drop locations is starting to look better. Just got the idea to modify flood lamps or street lights to encode the message in flickering. It can be obvious with timing looking like a broken light or imperceptible flickering like the old LED side channels.
This is still potentially insecure. One station might use it’s signaling ability to start running IPoMS (IP over Maritime Signaling), and end up getting hacked.