Blinding Attack on Self-Differencing APDs

System

Self-differencing single-photon detectors cancel the capacitive gating transient by subtracting the avalanche photodiode's output from a copy of itself delayed by one gate period. A threat actor (Eve) exploits that subtraction by sending light bright enough to trigger an essentially identical avalanche in every gate, so the circuit removes the real detections exactly as it removes the transient and the receiver (Bob) registers nothing [Jiang2013].

The avalanche photodiode is not driven out of its sensitive single-photon regime, which is what separates this from the better-known blinding attacks and why several of the standard detector defences do not fire. With the detector silenced, Eve intercepts the transmitter's (Alice's) states with her own copy of Bob's receiver and resends a brighter controlling pulse on top of the blinding train, which makes Bob record the bit she chose [Jiang2013,Gao2022].

She then listens to the ordinary public traffic Alice and Bob use for sifting, error correction and privacy amplification, and repeats the same steps; both sources state that this gives her the identical final key rather than merely the raw key. Both showed control of the detector in the laboratory, and neither ran the full attack on a QKD link. The calculated error rate stays where Alice and Bob expect it [Jiang2013,Gao2022].

Literature 7
[Gao2022] B. Gao, Z. Wu, W. Shi, Y. Liu, D. Wang, C. Yu, A. Huang, and J. Wu. "Ability of strong-pulse illumination to hack self-differencing avalanche photodiode detectors in a high-speed quantum-key-distribution system" (2022) 10.1103/PhysRevA.106.033713. arXiv:2205.04177.
[Jiang2013] M. Jiang, S. Sun, G. Tang, X. Ma, C. Li, and L. Liang. "Intrinsic imperfection of self-diferencing single-photon detectors harms the security of high-speed quantum cryptography systems" In: Phys. Rev. A 88, 062335. (2013) 10.1103/PhysRevA.88.062335.
[Jiang2024] Lianjun Jiang, Dongdong Li, Yuqiang Fang, Meisheng Zhao, Ming Liu, Zhilin Xie, Yukang Zhao, Yanlin Tang, Wei Jiang, Houlin Fang, Rui Ma, Lei Cheng, Weifeng Yang, Songtao Han, Shibiao Tang. "Countermeasure against blinding attack for single-photon detectors in quantum key distribution" (2024) 10.1088/1674-4926/45/4/042702. https://doi.org/10.1088/1674-4926/45/4/042702.
[KoehlerSidki2018] A. Koehler-Sidki, J. Dynes, M. Lucamarini, G. Roberts, A. Sharpe, Z. Yuan, and A. Shields. "Best-Practice Criteria for Practical Security of Self-Diferencing Avalanche Photodiode Detectors in Quantum Key Distribution" In: Phys. Rev. Appl. 9, 044027. (2018) 10.1103/PhysRevApplied.9.044027.
[KoehlerSidki2018a] A. Koehler-Sidki, M. Lucamarini, J. Dynes, G. Roberts, A. Sharpe, Z. Yuan, and A. Shields. "Intensity modulation as a preemptive measure against blinding of single-photon detectors based on self-differencing cancellation" In: Phys. Rev. A 98, 022327. (2018) 10.1103/PhysRevA.98.022327.
[Lee2016] M. Lee, B. Park, M. Woo, C. Park, Y. Kim, S. Han, and S. Moon. "Countermeasure against blinding attacks on low-noise detectors with a background-noise-cancellation scheme" In: Phys. Rev. A 94, 062321. (2016) 10.1103/PhysRevA.94.062321.
[Su2025] Jialei Su, Qingquan Peng, Jia-Lin Chen, Feng-Yu Lu, Zihao Chen, Junxuan Liu, De-Yong He, Shuang Wang, Anqi Huang. "Security analysis of orthogonal state attack on a high-speed quantum key distribution system" In: arXiv:2506.03718 [quant-ph]. (2025) 10.1038/s41534-026-01252-9. arXiv:2506.03718.
QID: A-0077
UUID: 5048c6a6-525e-4dd4-a178-46afcc8f0274
Tier: T0
Type: Quantum
Domain(s):
  • Quantum Communication
Tactic: Execution
Created: 2024-04-11
Updated: 2026-11-12
Maturity: Experimental
Technique properties

Attack type: Key extraction
Attack category: Detector-control attack
Target party: Receiver
Target function: Quantum layer
Quantum channel: Fibre optic, Free space
Security layers: L1 — Physical Layer (Quantum & Optical) L2 — Electronics & Analog Components
Maturity: Experimental
Operational properties

Technique → Countermeasures 4
Description Countermeasure

An upconversion receiver converts the signal before detection and detects it with a different detector, so no self-differencing InGaAs diode remains for this technique to act on.

CM-0024 Signal Upconversion

Bit-mapped gating randomises the mapping between detection events and key bits, so a blinded detector's clicks no longer correspond to a bit the threat actor can choose and her resends produce errors instead of a silent copy. It addresses the control step rather than the blinding step: this technique only yields a key because she resends her own detection results as bright controlling pulses, which is the step the randomised mapping breaks [Jiang2013]. Self-differencing detectors are gated, so the scheme can be deployed. No source tests it against the cancellation route, and the detector stays blind either way.

CM-0039 Bit-Mapped Gating

Prior hardware study is the modelling discipline, distinct from its own child CM-0052, which is the detector-specific characterisation including the capacitive-residual check. It is [KoehlerSidki2018]'s sixth criterion, model the behaviour of the detector, and why it matters here is measurable rather than rhetorical: the real blinding power sits more than three orders of magnitude above what a binary-detector model predicts, while a model with photocurrent feedback reproduced it. [Gao2022] shows the cost of treating a criterion derived from one hardware study as general, having satisfied the sub-50-kohm bias-resistor rule at 1 kohm and been blinded anyway.

CM-0086 Prior Hardware Component Study

Incorporating the imperfection into the security proof matters for this technique: [Jiang2013] concludes the loophole cannot be removed by redesigning the self-differencing detector, because it is the operating principle and not a circuit fault that the attack exploits. If the weakness cannot be engineered out, bounding the threat actor's information in the proof and paying the key-rate cost is one of the few paths left. The proof must also use the right detector model, since the standard blinding model assumes the diode leaves Geiger mode and [Jiang2013]'s measurement shows it does not.

CM-0014 Full Device Imperfections Incorporation in Scientific Security Proof
Technique → Security Layers 2
Description Category Impact Type

Physical layer, quantum and optical. The delivery is optical power arriving through the receiver's ordinary port far above single-photon level, and what the cancellation stage destroys is a physical avalanche that light created. Blinding onsets were measured on 200 MHz and 625 MHz detectors [Jiang2013,Gao2022].

L1 — Physical Layer (Quantum & Optical) Primary Target

Electronics and analog components, and for this technique the distinguishing layer. The failing element is the analog cancellation chain: a splitter, a one-gate delay, a recombiner and a discrimination comparator. [Jiang2013]'s measurement forces the assignment rather than suggesting it, since the diode is measured to stay in Geiger mode [Jiang2013,Gao2022].

L2 — Electronics & Analog Components Primary Target
See Also
A-0075 Detector Control by Exploiting Superlinearity
A-0076 Blinding Attack Using Dead Time
A-0078 Illuminating and Faked States Attack on SNSPDs
A-0080 Backflash Attack
A-0081 Spatial Misalignment Exploitation in QKD