IEEE Journal of Selected Topics in Signal Processing
Imprecise Probabilities Bounds for Quantum Key Distribution
Jorge Garcia-Diaz, Pino Caballero‐Gil, Daniel Escánez-Expósito
Abstract
Quantum Key Distribution generates a secret key between two parties with security rooted in quantum mechanics rather than computational assumptions. In practice, however, the channel noise and the eavesdropper's capabilities are never known with precision, an epistemic uncertainty that point-probability models do not capture. This work applies the theory of Imprecise Probabilities, through the notion of credal sets, to the security analysis of Quantum Key Distribution. It decomposes the observed Quantum Bit Error Rate into a benign and an adversarial contribution and, within a bounded-adversary model, bounds the eavesdropper's total induced disturbance across the two conjugate bases by a budget fixed by the protocol's abort threshold. The admissible bit and phase error contributions then form a coupled, non-rectangular credal set, a budget simplex, over which the lower prevision of the Devetak-Winter rate is evaluated in closed form. It is shown that the resulting worst case is attained at an interior allocation of the budget, strictly above the bound obtained by treating the two bases as independent intervals: the imprecise-probability treatment recovers secret key that no worst-case interval substitution can reproduce. This strict gap persists when no benign-noise floor is trusted and the entire observed error is attributed to the adversary, so the recovery does not rest on a trusted-noise premise; calibrating an in-laboratory floor is an optional refinement that further raises the rate and is itself treated as an imprecise quantity. The guarantee is robust to the absence of a precise calibration of the adversary and complements existing finite-key analyses by addressing structural rather than statistical uncertainty.