Slovakia Runs a Four-City Quantum Key Network Over Real Fiber — and Honestly Reports Where the Math Breaks

Entanglement-based QKD across 4 cities on deployed telecom fiber, with real-time key delivery. 13.5 bit/s over 109 km. The honest part: their privacy amplification overshoots finite-key bounds.

Imagine you and three friends each have a landline telephone, and you want to set up private conversations between every pair — but you only have one switchboard in the middle. The switchboard doesn't listen; it just sorts calls by the pitch of each person's voice, routing each pair onto their own private frequency. Now imagine that instead of voices, you're splitting entangled photons by color, and instead of privacy being social convention, it's enforced by quantum mechanics. That's the core mechanism here: a central source of polarization-entangled photon pairs, passively split by wavelength into dedicated city-to-city channels, with no trusted node required between the three outer cities. The committed claim: this is the first deployed, simultaneously operated, entanglement-based QKD network connecting four cities over existing telecom fiber, with keys delivered in real time via the ETSI standard interface — not just coincidence rates reported after the fact. The network spans western Slovakia (Bratislava, Banská Bystrica, Žilina, Zvolen), with fiber lengths up to 142.9 km and accumulated losses up to 32.1 dB. The topology is a passive wavelength-routed mesh for three cities, with the fourth connected through the central node. The authors also introduce a matrix formalism for designing these wavelength-routing topologies algorithmically, which is a useful engineering contribution separate from the demonstration itself. The numbers that matter: 13.5 bit/s of secure key over the 109 km Bratislava–Žilina link after full post-processing, and 4.3 bit/s over the hardest link (Bratislava–Banská Bystrica, 32.1 dB loss, no direct fiber). These are real delivered key rates, not theoretical projections from raw coincidence counts, which puts this paper ahead of most QKD demonstrations that stop at 'we could have generated keys if we'd run the full stack.' The 142.9 km link (Bratislava–Zvolen) produced no positive finite-key rate under the implemented scheme — a result the authors report honestly rather than burying. The integrity story is where this paper earns real respect. The authors benchmark their implemented heuristic privacy amplification against rigorous finite-key security bounds and find that the heuristic sometimes exceeds what the finite-key analysis guarantees — meaning the keys are potentially shorter than what strict security requires. They don't hide this. They show that replacing their detect-and-discard error handling with proper error correction would boost finite-key rates by 4.1–6.6×, and would rescue the 142.9 km link from zero to positive key generation. This is unusually honest self-assessment for a field where demonstrations tend to emphasize the best-case numbers. Architecturally, this sits in the entanglement-based (BBM92-family) QKD lineage, which is harder to deploy than prepare-and-measure (BB84) but offers device-independence advantages in principle. The passive wavelength routing via CWDM filters is the engineering enabler — no active switching, no trusted relay nodes between the three outer cities. The partial non-local chromatic dispersion compensation is the optical trick that makes the longer links viable. The matrix formalism for topology design is a nice contribution for network planners but not the headline result. The field context: QKD networks are transitioning from lab demonstrations and point-to-point links to deployed multi-node infrastructure. China's Beijing–Shanghai backbone, the Cambridge and Bristol networks in the UK, and various European testbeds are all pushing this frontier. Most rely on trusted nodes (which are a security assumption, not a physics guarantee) or prepare-and-measure protocols. Entanglement-based networks without trusted nodes are rarer and harder. This paper advances that specific frontier on real infrastructure, but the key rates (single-digit to low-double-digit bit/s) remain orders of magnitude below what's needed for practical high-bandwidth encryption. The honest bottom line: this is a well-executed infrastructure demonstration with unusually transparent security accounting. It doesn't claim to solve QKD networking — it shows what works, what doesn't, and where the concrete improvements lie. The 4.1–6.6× improvement from proper error correction is not speculative; it's a calculable upgrade path. The field needs more papers like this that report real delivered rates and honest security gaps, rather than optimistic projections from raw data.