Sharp finite statistics for quantum key distribution

TL;DR

This paper introduces a new, highly precise statistical method for quantum key distribution that improves security analysis, reduces necessary block sizes, and enhances confidence interval estimation for nonidentical Bernoulli parameters.

Contribution

It presents an unprecedentedly tight statistical bound for QKD, outperforming traditional hypergeometric tail bounds and enabling more efficient and accurate security assessments.

Findings

New tight statistical bounds for QKD security analysis

Improved confidence intervals for Bernoulli parameters

Reduced minimum block sizes for secure QKD implementation

Abstract

The performance of quantum key distribution (QKD) heavily depends on statistical inference. For a broad class of protocols, the central statistical task is a random sampling problem, customarily addressed using a hypergeometric tail bound due to Serfling. Here, we provide an alternative solution for this task of unprecedented tightness among QKD security analyses. As a by-product, confidence intervals for the average of nonidentical Bernoulli parameters follow too. These naturally fit in statistical analyses of decoy-state QKD and also outperform standard tools. Last, we show that, in a vast parameter regime, the use of tail bounds is not enforced because the cumulative mass function of the hypergeometric distribution is accurately computable. This sharply decreases the minimum block sizes necessary for QKD, and reveals the tightness of our analytical bounds when moderate-to-large blocks are considered.