Is there actual research into programming on untrusted hardware? I mean, I gotta think that if you think any part of your computer might be compromised you can't trust it at all.
Because a communications link is generally only a passive part of the system, it relays the signal and that is all...
Computing on untrusted hardware is like trying to convert a plaintext secret message to ciphertext while an enemy agent with perfect memory is watching over your shoulder, not exactly easy.
Because encryption hides the data content, whereas you need it in the clear to computationally manipulate it in a useful way, at least with conventional computational frameworks. How is a program going to make a decision at a control flow point if the data it needs to make the decision is encrypted? And if the data's not encrypted, and you're running on untrusted hardware, then it's compromised.
