I built a CPU myself out of 74xx logic, but I can't be sure that every input on every chip isn't also connected internally to an integrated computer that stores everything it sees and can transmit it to a van at 20 metres range (easily enough for someone to drive up to my house and retrieve the contents periodically).
The attacker can then reconstruct everything that happened on the CPU and work out what I've been up to. Or, more likely, throw away almost everything and only look at what was printed to the console, because everything interesting goes to the console anyway.
In fact this would be easier-than-average to achieve because the clock speed is low, I don't run the machine for very long or very often, and the density of actual electronics to "free space" on the ICs is very low (more space for shenanigans), and all my schematics are on my github project so even though you don't know which chip is connected to which other, you shouldn't have too much trouble deriving it.
So what now? Do we just give up? Is all computing fundamentally impossible? Or do we accept that nothing is perfect and just get on with it?
The attacker can then reconstruct everything that happened on the CPU and work out what I've been up to. Or, more likely, throw away almost everything and only look at what was printed to the console, because everything interesting goes to the console anyway.
In fact this would be easier-than-average to achieve because the clock speed is low, I don't run the machine for very long or very often, and the density of actual electronics to "free space" on the ICs is very low (more space for shenanigans), and all my schematics are on my github project so even though you don't know which chip is connected to which other, you shouldn't have too much trouble deriving it.
So what now? Do we just give up? Is all computing fundamentally impossible? Or do we accept that nothing is perfect and just get on with it?