Non-asymptotic security completion of measurement-device-independent quantum secure direct communication: finite-key secrecy capacity, public-randomization leakage, and device-level characterization of the gain gap

Measurement-device-independent quantum secure direct communication (MDI-QSDC) lets two users exchange confidential messages directly over a quantum channel while an untrusted middle party performs all quantum measurements, removing detector side channels. The existing quantitative security analysis (Niu, Wu, Yin and Long, 2020) gives only asymptotic lower bounds on the secrecy capacity, obtained by reducing coherent attacks to collective attacks via a public randomized permutation and Bell-diagonalizing local unitaries, and it introduces an abstract gain-gap parameter relating the legal users' and eavesdropper's detection gains without deriving it from device physics. The open problem is to complete this analysis to an implementation-grade security statement: a finite-blocklength secrecy-capacity lower bound with explicit finite-size corrections, a precise accounting of the information leaked by publishing the randomization indices, and a physically grounded determination of the gain gap under realistic attack models.

Background

Quantum secure direct communication (QSDC) transmits a confidential message directly through a quantum channel, without first distributing a secret key and then encrypting; security comes from quantum-mechanical disturbance of eavesdropping rather than from computational assumptions. Two foundational protocols are the two-step protocol, based on blocks of Einstein-Podolsky-Rosen pairs (shared entangled particle pairs), and the DL04 protocol (Deng and Long, 2004), in which Alice encodes each message bit onto a single photon by one of four unitary operations (a 'quantum one-time pad') and returns it to Bob. A central quantity is the secrecy capacity $C_S$, the largest rate at which Alice can convey the message to Bob while keeping Eve's information negligible; by the Csiszár-Körner wiretap theory it is lower-bounded by the difference of mutual informations $I(A:B)-I(A:E)$ between the legal parties and the eavesdropper. Measurement-device-independent QSDC (MDI-QSDC), proposed by Niu et al. (2018), removes the most exploitable practical weakness — detector side channels — by having Alice and Bob each send photons to an untrusted middle node (Charlie) who performs Bell-state measurements (joint projections onto maximally entangled two-qubit states); whatever detector imperfections Charlie has, they cannot leak message information because the measurement outcome, not the device, is authenticated by correlations. Niu, Wu, Yin and Long (2020) completed the first quantitative security analysis of two MDI-QSDC protocols (the MDI version of the two-step protocol and of DL04). Their proof reduces general coherent attacks (attacks that act jointly across all transmitted signals) to collective attacks (attacks acting independently on each signal) by an asymptotic quantum de Finetti argument (the quantum analogue of the classical theorem that a permutation-invariant sequence of systems is close to a mixture of independent, identically distributed states): Alice and Bob publicly apply a randomized permutation $\mathcal{P}_\mu$ (index $\mu$ announced on the public channel) and local random unitaries $\mathcal{U}_\nu$ (index $\nu$ public) that erase coherences and render the shared two-qubit state Bell-diagonal with parameters constrained by the observed bit and phase error rates. Using the Holevo bound (an upper bound on the accessible information carried by a quantum state), they bound Eve's information and obtain secrecy-capacity lower bounds of the form $C_S \ge Q\{2-H(\mathbf{E})-\eta[h(\epsilon_z)+h(\epsilon_x)]\}$ for the entanglement-based protocol and $C_S \ge Q[1-h(e)-\eta h(\epsilon_u)]$ for the single-photon protocol, where $Q$ is Bob's detection gain, $h$ is the binary entropy function, $H(\mathbf{E})$ is the Shannon entropy of the observed error distribution, the $\epsilon$'s are estimated error rates, and $\eta$ is an abstract 'gain gap' between the A–E and A–B channels, bounding how much more of the traffic Eve can collect than Bob; in the convention of the field's survey it is the ratio $\eta = Q_{\mathrm{Eve}}/Q_{\mathrm{Bob}}$ of Eve's effective reception gain to Bob's detection gain. Four gaps in this analysis are acknowledged in the paper or left untreated. First, the de Finetti reduction is asymptotic, so no finite-blocklength (finite-key) bound or finite-size correction term exists, while real systems transmit finitely many blocks. Second, the randomization indices $\mu$ and $\nu$ travel on the public channel and could leak information; the paper notes its formulas remain lower bounds only if these operations are removed, and quantifies nothing about the leakage they cause when kept. Third, $\eta$ is a free parameter (set to 1 only in the idealized depolarizing-channel simulation, where the channel maps any input to a mixture with the maximally mixed state), with no derivation from device characteristics under realistic attacks such as detector-efficiency side channels or loss-dependent eavesdropping. Fourth, performance was illustrated only by depolarizing-channel simulations; no end-to-end experiment has validated the predicted secrecy-capacity curves for practical devices with multi-photon sources and detector imperfections. Finite-key analysis is a mature technology in quantum key distribution (e.g., via Rényi-entropy and concentration-based corrections), so the question is whether it can be adapted to MDI-QSDC's distinctive structure: two-way block transmission, quantum memory, and an untrusted middle node whose Bell-state measurements implement entanglement swapping (the measured photons are destroyed and the two remote photons previously paired with them become entangled).

Problem Statement

Complete the quantitative security analysis of the measurement-device-independent quantum secure direct communication protocols (the MDI two-step and MDI-DL04 protocols of Niu et al.) so that it yields an implementation-grade secrecy statement, replacing the asymptotic i.i.d. framework of Niu, Wu, Yin and Long (2020) by a non-asymptotic one and eliminating its unquantified gaps. Concretely, the open problem is to determine the secrecy capacity of these protocols under finite-length effects and practical system parameters, which requires closing the following gaps of the 2020 analysis: (1) derive a finite-blocklength (finite-key) lower bound on the secrecy capacity of these protocols — in the standard $\varepsilon$-security sense used in finite-key quantum key distribution and in finite-blocklength QSDC analysis (reference 9), i.e., an achievable message rate that is $\varepsilon$-secret for an explicit security parameter $\varepsilon$ — with finite-size correction terms explicit in the block length, so that security against general coherent attacks holds at finite block length, in place of the asymptotic quantum de Finetti reduction of the 2020 analysis; (2) give a precise information-theoretic accounting of the leakage caused by the public randomization steps — the published permutation index $\mu$ and local-unitary index $\nu$ used for Bell-diagonalization — either bounding the resulting capacity penalty within the same secrecy-capacity statement, or exhibiting a modification that removes the leakage, in which case the secrecy-capacity bound is that of the explicitly identified modified protocol; and (3) characterize the gain-gap parameter $\eta$ physically, so that the capacity bound becomes computable from measured device data: derive $\eta$, or explicit bounds on it valid under stated realistic attack models, from experimental device characteristics (for example, detector efficiencies and their mismatch, dark counts, loss, and multi-photon emission — these are illustrative of the device parameters to be modelled, not an exhaustive checklist), where the attack models cover effects such as loss-dependent eavesdropping and side-channel or efficiency-mismatch effects at the legitimate users' own devices (their sources and local security-check measurements; Charlie's measurement devices are untrusted by the MDI premise), or identify a protocol modification that provably fixes the effective gain gap.

References

1. Peng-Hao Niu, Jia-Wei Wu, Liu-Guo Yin, Gui-Lu Long, 'Security analysis of measurement-device-independent quantum secure direct communication', Quantum Information Processing 19, 356 (2020). DOI: 10.1007/s11128-020-02840-0; arXiv:2006.07184, https://arxiv.org/abs/2006.07184
2. F. G. Deng, G. L. Long, 'Secure direct communication with a quantum one-time pad', Physical Review A 69, 052319 (2004). DOI: 10.1103/PhysRevA.69.052319, https://doi.org/10.1103/PhysRevA.69.052319
3. Peng-Hao Niu, Zeng-Rong Zhou, Zai-Sheng Lin, Yu-Bo Sheng, Liu-Guo Yin, Gui-Lu Long, 'Measurement-device-independent quantum communication without encryption', Science Bulletin 63(20), 1345-1350 (2018). DOI: 10.1016/j.scib.2018.09.009, https://doi.org/10.1016/j.scib.2018.09.009
4. Nayana Das, Goutam Paul, 'Improving the security of "Measurement-Device-Independent Quantum Communication without Encryption"', Science Bulletin 65(24), 2048-2049 (2020). DOI: 10.1016/j.scib.2020.09.015; arXiv:2006.05263, https://arxiv.org/abs/2006.05263
5. Zhang-Dong Ye, Dong Pan, Zhen Sun, Chun-Guang Du, Liu-Guo Yin, Gui-Lu Long, 'Generic security analysis framework for quantum secure direct communication', Frontiers of Physics 16, 21503 (2021). DOI: 10.1007/s11467-020-1025-x, https://doi.org/10.1007/s11467-020-1025-x
6. Jiawei Wu, Gui-Lu Long, Masahito Hayashi, 'Quantum secure direct communication with private dense coding using a general preshared quantum state', Physical Review Applied 17, 064011 (2022). DOI: 10.1103/PhysRevApplied.17.064011, https://doi.org/10.1103/PhysRevApplied.17.064011
7. Xiang-Jie Li, Dong Pan, Gui-Lu Long, Lajos Hanzo, 'Single-Photon-Memory Measurement-Device-Independent Quantum Secure Direct Communication—Part II: A Practical Protocol and its Secrecy Capacity', IEEE Communications Letters 27(4), 1060-1064 (2023). DOI: 10.1109/LCOMM.2023.3247176; arXiv:2212.05661, https://arxiv.org/abs/2212.05661
8. Ze-Zhou Sun, Dong Pan, Dong Ruan, Gui-Lu Long, 'One-Sided Measurement-Device-Independent Practical Quantum Secure Direct Communication', Journal of Lightwave Technology 41(14), 4680-4690 (2023). DOI: 10.1109/JLT.2023.3244880, https://doi.org/10.1109/JLT.2023.3244880
9. Ze-Zhou Sun, Dong Pan, Yuan-Bin Cheng, Yu-Chen Liu, Dong Ruan, Gui-Lu Long, 'Multi-Intensity Quantum Secure Direct Communication Relying on Finite Block-Length', IEEE Transactions on Communications 72(8), 4633-4647 (2024). DOI: 10.1109/TCOMM.2024.3375814, https://doi.org/10.1109/TCOMM.2024.3375814
10. Ze-Zhou Sun, Yuan-Bin Cheng, Dong Ruan, Dong Pan, 'Single-photon measurement-device-independent quantum secure direct communication', Optics Communications 562, 130745 (2024). DOI: 10.1016/j.optcom.2024.130745, https://doi.org/10.1016/j.optcom.2024.130745
11. Dong Pan, Gui-Lu Long, Liuguo Yin, Yu-Bo Sheng, Dong Ruan, Soon Xin Ng, Jianhua Lu, Lajos Hanzo, 'The Evolution of Quantum Secure Direct Communication: On the Road to the Qinternet', IEEE Communications Surveys & Tutorials 26(3), 1898-1949 (2024). DOI: 10.1109/COMST.2024.3367535; arXiv:2311.13974, https://arxiv.org/abs/2311.13974
12. Peng-Hao Niu, Fei-Hao Zhang, Rui-Song Bao, Gui-Lu Long, 'Link optimization of duplex quantum communication systems', Science China Information Sciences 69 (2026). DOI: 10.1007/s11432-025-4611-8, https://doi.org/10.1007/s11432-025-4611-8