High-rate multipartite quantum secret sharing with composable security

AI summary

Overview

Research area: Experimental quantum cryptography and quantum communication networks — specifically multipartite quantum secret sharing (QSS) using photonic GHZ states.

Technical level: Intermediate. The paper is experimentally oriented but relies on a composable finite-key security framework and entropic uncertainty relations that require some familiarity with quantum key distribution theory.

Scope: The paper reports an experimental implementation and evaluation of a discrete-variable $(n,n)$-threshold quantum secret sharing protocol with composable security against general attacks (including participant attacks), using a high-rate four-photon GHZ source and a 20 km four-arm fibre star network.

What This Paper Is About

Quantum secret sharing lets a dealer split a secret among $n$ participants so that only authorised groups can reconstruct it, with quantum mechanics providing eavesdropping detection. The canonical GHZ-based protocol (Hillery–Bužek–Berthiaume) is vulnerable to "participant attacks," in which dishonest insiders exploit the public certification step — for example by measuring last — to pass checks while learning the secret. This paper experimentally evaluates a recently developed discrete-variable $(n,n)$-QSS protocol whose finite-key analysis provides composable security against such attacks, and benchmarks how fast such a system can actually run over realistic fibre.

Key Contributions

1. Experimental demonstration of a composably secure $(n,n)$-QSS protocol. The authors implement the discrete-variable protocol of Ref. [51] with photonic four-qubit GHZ states, where security against arbitrary unauthorised coalitions is derived from pairwise dealer–participant correlations via entropic uncertainty relations.

2. A high-rate GHZ source benchmark. Two domain-engineered entangled photon pair sources (aperiodically-poled KTP crystals in Sagnac interferometers) produce four-qubit GHZ states at rates above $5\times10^{3}$ fourfold events per second, with a measured maximum of 5025 counts/s and a maximum asymptotic secret key rate of $750 \pm 10$ bits per second.

3. Network-scale evaluation with finite-key analysis. The state is distributed through a four-arm star network with 20 km of fibre in total (5 km links, approximately 5 dB loss), collecting 22 million rounds over 10 hours of laboratory time, and the finite-key bound is projected to a randomised 24-hour run.

4. A candid accounting of practical gaps. The paper quantifies the benefit of basis biasing ($p_{\rm key}^{*} = 0.947$ for the complete dataset), analyses multiphoton filtering with pseudo-photon-number-resolving detectors (a roughly three-fold key-rate gain that is nonetheless insecure with threshold detectors), and explicitly states that blockwise basis acquisition does not yet constitute a fully cryptographic implementation.

Main Findings

• Asymptotic key rate: The maximum secure key rate measured was $750 \pm 10$ bits/s at 600 mW pump power, where the source produced an average fourfold rate of $r = 3970$ counts/s with $Q_X = 0.082 \pm 0.001$ and $\max[Q_{A,B_i}] = 0.0808 \pm 0.0004$, giving an average asymptotic key rate (AKR) of $0.188 \pm 0.003$.

• State fidelity versus rate trade-off: At 680 mW the four-qubit GHZ fidelity was $0.842 \pm 0.002$ at 5025 counts/s; at 100 mW it was $0.948 \pm 0.018$ at 111 counts/s. Full tomography at 100 mW (112 fourfolds/s, $3.3\times10^{3}$ rounds per basis, $2.7\times10^{5}$ rounds total) gave a purity of $0.919 \pm 0.013$.

• Source characterisation: Individual Bell state fidelities were $0.991 \pm 0.003$ and $0.990 \pm 0.003$ at 20 mW, extrapolating to $0.928 \pm 0.005$ at 600 mW; dependent two-photon interference visibilities were $0.981 \pm 0.002$ and $0.992 \pm 0.002$; fusion-gate visibility was $0.960 \pm 0.003$ at 20 mW, falling to $0.837 \pm 0.003$ at 600 mW.

• Fibre network performance: Over 20 km with roughly 5 dB total loss, the fourfold rate was $r = 1197$ per second with $Q_X = 0.090 \pm 0.003$ and $\max[Q_{A,B_i}] = 0.075 \pm 0.001$.

• Finite-key projection: A randomised 24-hour execution with optimised basis probability yields a composable finite-key lower bound of 8.7 Mbits, versus $7.4\times10^{5}$ bits with the symmetric basis probability ($p_{\rm key} = 0.5$). A positive key is obtained after 6.5 minutes, using $4.7\times10^{5}$ total rounds.

• Basis biasing recovers QKD-like efficiency: Because participants are untrusted, bases cannot be pre-shared, and symmetric choices limit the key-generation probability to $\eta_{\rm key} = 1/2^{n}$. Biasing reaches the same asymptotic efficiency as trusted QKD protocols, and the optimum bias approaches $p_{\rm key}^{*} = 0.947$ for the full dataset.

• Multiphoton filtering: Discarding double-click events at 600 mW reduced $Q_x$ from 0.080 to 0.035 (a 56% decrease) and increased the AKR by approximately 250%, with an extrapolated filtered SKR maximum of 2700 bits/s at 900 mW. This is a 3-fold improvement but cannot be used securely with threshold detectors, owing to a known side-channel in which an adversary controls which rounds are discarded.

• Explicit security caveat: The motorised analysers switch too slowly for per-round basis choice, so settings were recorded in blocks (200 blocks of roughly 110,000 rounds each, using 12 of the 16 measurement settings). The finite-key numbers are therefore a performance projection under the assumption that source, channels and measurement apparatus remain stationary, not a full cryptographic execution.

Methodology in Plain English

The team built a photonic source that creates entangled photon pairs and then entangles two such pairs together. A 775 nm picosecond pulsed laser (80 MHz repetition rate, 1.3 ps pulses) pumps two aperiodically-poled KTP crystals inside Sagnac interferometers, producing photon pairs at 1550 nm. One photon from each source interferes on a polarising beamsplitter; when one photon exits each output port, the four photons are post-selected into a GHZ state. The gate succeeds with probability one-half.

Each of the four photons is sent through fibre to a measurement station with polariser analysers and superconducting nanowire single-photon detectors. Users measure in the $Z$ or $X$ basis. Rounds where everyone measured $X$ become key-generation rounds; rounds where the dealer Alice and at least one participant measured $Z$ become check rounds, from which pairwise error rates $Q_{A,B_i}$ are computed to bound what any unauthorised coalition or eavesdropper could know. The remaining combinations are discarded. The team first benchmarked the source with no transmission loss across pump powers from 100 mW to 680 mW, then deployed it in a star network of four 5 km fibre spools (losses 0.19–0.21 dB/km) with fibre polarisation controllers correcting random unitary rotations. Finally, they applied the finite-key analysis to the measured error statistics, treating the measured data as if it had been gathered under genuine random basis selection at various bias settings.

Why This Matters

Impact on research: Most quantum cryptography experiments are point-to-point. This work demonstrates a genuinely multipartite cryptographic primitive — one GHZ state coordinates all users in a single network use rather than $n$ separate links — and pairs it with a composable finite-key security claim that explicitly includes dishonest participants in the adversary model. It also sets a concrete key-rate and fidelity benchmark for four-photon GHZ experiments.

Real-world applications:

• Secure multiparty computation and threshold cryptography, where no single party should hold the whole secret.
• Controlled access to information and distributed storage of sensitive data across custodians, replacing classical schemes such as Shamir's with eavesdropping detection.
• Future quantum network services in star or bottleneck topologies where multipartite correlations use network resources more efficiently than pairwise links.
• Quantum state sharing, the natural extension to distributing an unknown quantum state to an authorised group.

Industry relevance: The results give network architects a realistic sense of achievable QSS rates (hundreds of bits per second over 20 km with current SPDC sources) and identify the specific engineering bottlenecks — multiphoton emission, basis-switching speed, and detector type — that determine whether deployed systems can move from proof of principle to cryptographic operation.

Future Directions

• Closing the basis-selection gap: Implementing fast active basis switching, or a properly analysed passive implementation using biased beamsplitters, so that the finite-key projections become an actual cryptographic guarantee rather than a stationarity-dependent projection, followed by real information reconciliation and privacy amplification.
• Extending to general access structures: The $(n,n)$ threshold is the simplest case; extending composable security proofs to graph-state-based $(k,n)$ schemes would allow flexible authorised subsets and protect against dishonest participants in richer network topologies.
• Better sources and detectors: Near-deterministic emitters (quantum dot sources have shown heralded three-photon GHZ generation at approximately 500 counts/s with fidelities around 0.72), time-multiplexed probabilistic sources, boosted fusion gates, and photon-number-resolving detectors placed after polarisation measurement to enable secure multiphoton filtering.
• Quantum state sharing and scaling: Moving from classical secrets with quantum security to sharing unknown quantum states, and using quantum repeaters or post-matching approaches to extend multipartite QSS beyond short distances and few users.

Target Audience

Experimental quantum communication researchers; quantum cryptography theorists interested in composable security for multipartite protocols; photonics engineers working on entangled photon sources and GHZ state generation; and quantum network architects assessing the near-term performance of multiparty cryptographic primitives. Readers without a background in quantum key distribution security proofs will find the experimental sections accessible, but the finite-key analysis in the appendices assumes familiarity with smooth min-entropy, Hoeffding and Serfling bounds, and epsilon-security budgets.

Authors’ abstract

Future quantum communication networks will conceivably support cryptographic tasks that require entanglement among more than two users. Quantum secret sharing is a prime example where entanglement provides a direct means to coordinate untrusted parties with security from eavesdropping in a multi-party setting. However, the canonical GHZ-based protocols can be vulnerable to participant attacks, in which untrusted parties try to learn the secret without collaborating. Here, we experimentally evaluate a discrete-variable $(n,n)$-threshold quantum secret-sharing protocol whose finite-key analysis provides composable security against general attacks, including participant attacks. Using two domain-engineered entangled photon pair sources, we generate 4-qubit GHZ states at rates above $5\times10^3$ fourfold events per second and a maximum asymptotic secret key rate of $750 \pm 10$ bits per second. We then distribute the state through a 4-arm star network comprising 20km of fibre in total. From the measured event rates and error statistics, we infer that a randomised 24-hour execution with the optimised basis probability would yield a composable finite-key lower bound of 8.7 Mbits, under the assumption that the measured source and device statistics remain stationary.