Improving the secret key rate without changing the experimental setup by McDiarmid inequality

Abstract

The foci of quantum key distribution researches are making it practical, reliable, error-tolerant and most importantly fast. To achieve a high provably secret key rate in practical situation with a finite raw key length, most works to date focus on either better protocol, better experiments or better theories to deal with issues of imperfect apparatus. Is it possible that the secret key rate of a current experiment is actually higher than the one given by existing proofs? Here I show that this is indeed the case in the finite raw key length situation. I first point out why existing proofs are ineffective in this aspect; and then I introduce powerful method from classical statistics community to tackle the problem. In particular, I show how to use McDiarmid inequality to increase the provably secure key rate of the BB84 scheme with decoy through a realistic optical fiber channel by at least 30% without changing existing experimental setup and classical processing procedure. Details of this work can be found in arVix:1806.05073.