Intel OEM Private Key Leak: A Blow to UEFI Secure Boot Security

Unless and until we get to efficient homomorphic compute, these measures will only ever be security via obscurity.

I don't see how homomorphic encryption is particularly applicable to secureboot.

You want to be able to deploy and execute code outside the control of whoever physically controls the machine.

Either you implement it with security features hidden from the device holder, in which case it will always be broken eventually, or you guarantee the capabilities with mathematics - in which case a security break cannot happen even if the physical machine's description is completely public.

There are certainly layers to this that I'm missing, but I think homomorphic compute is the only unbreakable answer to secure compute in general.

My understanding (which might be wrong, crypto is a complex topic and i am an amateur) is that homomorphic would hide the data being worked on from the algorithm working on it. Here we want to verify the (non-secret) algorithm has been approved (code signing) which we then run on non-secret data. I don't think homomorphic encryption can help with that since its kind of a different problem.

The issue here, of the key holder leaking the key, also seems impossible to work around in general, since the requirements are: 1) there exists someone who can sign code. 2) that person cannot screw up (e.g. leak the key) and allow the wrong code to be signed. These are pretty contradictory requirements, that no amount of crypto can fix. Ultimately it is a social problem not a technical one; there is no full technical definition of misusing a key. There are things that can help - like HSMs, splitting the key between multiple parties, having better methods of revoking and replacing compromised keys (hard without network access and an unwillingness to brick old devices). Not the same domain, but AACS is an interesting example of a system somewhat resiliant to key compromise.