Who can actually read it? The provider's own log, live

server.py is byte-for-byte the same code every time — it just print(f"SERVER SAW: {payload}")s whatever it's handed. What changes is only what Alice chooses to send. Pick a message, flip the mode, and read the provider's own log for yourself — same math a real RSA key-wrap and AES-style cipher would do, just toy-sized so every step fits on screen.

Mode — which docker-compose file would be running:

Malicious server substitutes its own key for Bob's (the root-of-trust gap — not what these two containers actually demonstrate; only meaningful once Mode is Fixed)

What Alice hands the server, and the server's own log line for it:

What actually got computed (toy-sized RSA key-wrap + an AES-style cipher, so every number fits on screen — real modular exponentiation, real tag check, nothing scripted):

Who can actually read it? The provider's own log, live

The operator — reading its own log

Bob — decrypting with his real private key