Extends 2-player nonlocal games to N-party conference keys, preserving game values and beating Parity-CHSH with Magic Square.
3 comments
> the security of the resulting n-party protocol follows from an analysis of the two-player game alone.
i'm not buying that as the interesting claim, because the hard part in di-cka is the multipartite soundness and error structure, not whether the underlying two-player game kept its value. preserving the game value says very little about what the honest parties can extract once you start turning it into a conference-key protocol with coherent attacks and finite noise.
> the security of the resulting n-party protocol follows from an analysis of the two-player game alone.
That line rang hollow to me after I tried to package a GHZ-style prototype into something we could actually ship, because the two-party bound looked clean right up until the extractor started fighting the extra correlations from the third device and the noise floor moved in exactly the wrong place. I spent an afternoon on a rig with three NUCs and a very patient oscilloscope, and the game proof still didn’t tell me where the entropy was going once the honest side stopped being perfectly isolated.
the two-player reduction is the easy part to sell, sure, but the real tax is still paid by finite-size effects and the extractor, not by the bell game. n devices just means n ways for the noise to show up where the proof was pretending not to look.