FMD + TEE for private state sync proposal

[kike]

Hi @vveiln . Thanks for your replies!

@vveiln

I think it would be helpful to explicitly say what the limits of the first iteration are and what is the ideal functionality we want to achieve after n iterations.

I wrote the basic proposal based on this comment:

@cwgoes

Most important aspect of first proposal is a single detection layer that can run FMD in a TEE with the right cryptography and some proof that the clue is correct.

Regarding what we ideally want to achieve, yes that’s something to think about it.

@vveiln

@kike

I think that at least the last TEE node of the decryption layer should communicate with users directly using a confidential and authenticated channel (If there is only the detection layer, then the last detection TEE.)

I wonder if we can make the whole system stateless so that the tees didn’t have to be aware of each other’s work or existence. I feel like we add a lot of complexity by having it this way. I don’t have an answer to that, but it is something to think about

I was answering how users and TEEs communicate. I think we agree on this.

You may be right, but I also think we may benefit by having TEEs working together, specially for privacy.

Re the ZKPs of FMD falg consistency. Sorry it wasn’t very clear, hopefully in this post is clearer.