Injection-Locking Attack (ILA)
In the Injection-Locking Attack (ILA) the threat actor (Eve) sends her own laser light backwards down the quantum channel into the transmitter's (Alice's) laser source. The injected light forces Alice's laser to follow Eve's, which changes the light Alice emits in a way that depends on the setting she just chose, so Eve can read that setting from Alice's own outgoing pulses without stopping the running session [Pang2020,Zhang2022].
There are two routes: one demonstrated in the laboratory [Pang2020], one analysed from measured laser spectra and simulation [Zhang2022]. Where Alice encodes in polarisation, her own polarisation optics let Eve's light in fully only when Eve's choice matches Alice's, so a shift in Alice's emitted frequency tells Eve which polarisation Alice used [Pang2020]. Where Alice's laser has no isolator at all, which some twin-field and phase-encoded systems require by design, Eve instead switches her injection between two brightness levels and reads whether each pulse was a signal or a decoy [Zhang2022].
Either way Eve ends up holding part of the secret key, and the leftover information also opens the door to a follow-on attack such as intercept-and-resend or photon-number splitting [Pang2020,Zhang2022].
| [Lovic2023] | V. Lovic, D. G. Marangon, P. R. Smith, R. I. Woodward, and A. J. Shields. "Quantified effects of the laser-seeding attack in quantum key distribution" In: Phys. Rev. Appl., vol. 20, no. 4, p. 044005. (2023) 10.1103/PhysRevApplied.20.044005. |
| [Pang2020] | X. Pang, A. Yang, C. Zhang, J. Dou, H. Li, J. Gao, and X. Jin. "Hacking Quantum Key Distribution via Injection Locking" In: Phys. Rev. Appl. 13, 034008. (2020) 10.1103/PhysRevApplied.13.034008. |
| [Ponosova2022] | A. Ponosova, D. Ruzhitskaya, P. Chaiwongkhot, V. Egorov, V. Makarov, and A. Huang. "Protecting Fiber-Optic Quantum Key Distribution Sources against Light-Injection Attacks" In: PRX Quantum, vol. 3, no. 4, p. 040307. (2022) 10.1103/PRXQuantum.3.040307. |
| [Ponosova2025] | Anastasiya Ponosova, Irina Zhluktova, Daria Ruzhitskaya, Daniil Trefilov, Anqi Huang, Alexey Wolf, Vladimir Kamynin, Vladimir Tsvetkov, Vadim Makarov. "Pulsed laser attack at 1061 nm potentially compromises quantum key distribution". (2025) arXiv:2506.19091. |
| [Wiesemann2024] | Jerome Wiesemann, Fadri Grunenfelder, Ana Blazquez, Nino Walenta, Davide Rusca. "Evaluation of quantum key distribution systems against injection-locking attacks" (2024) 10.1063/5.0260685. arXiv:2412.10290. |
| [Zhang2021] | G. Zhang, I. W. Primaatmaja, J. Y. Haw, X. Gong, C. Wang, and C. C. W. Lim. "Securing Practical Quantum Communication Systems with Optical Power Limiters" (2021) 10.1103/PRXQuantum.2.030304. https://link.aps.org/doi/10.1103/PRXQuantum.2.030304. |
| [Zhang2022] | X. Zhang, M. Jiang, Y. Wang, Y. Lu, H. Li, C. Zhou, Y. Zhou, and W. Bao. "Analysis of an injection-locking-loophole attack from an external source for quantum key distribution" In: Phys. Rev. A 106, 062412. (2022) 10.1103/PhysRevA.106.062412. |
f9044361-179d-4ee5-b30b-41f2ff593520
- Quantum Communication
Technique properties
Operational properties
| Description | Countermeasure |
|---|---|
|
Wiesemann2024 states the two branches open to a decoy-state BB84 security proof. Current security proofs either assume uniform phase randomization of Alice's signals, which is compromised by practical limitations and attacks like injection locking, or rely on a partially characterized phase distribution [Wiesemann2024]. CM-0014 is that second branch, and this technique is the reason the first is unavailable. The same work supplies what the branch needs in practice, a measurable metric for how far the phase distribution has been pulled plus methods that are source-agnostic and can be used to evaluate general QKD systems against injection-locking attacks, including in evaluation laboratories through black-box testing [Wiesemann2024]. The control does not stop the injection and does not detect it, it bounds what the resulting de-randomisation can leak and hands that bound to privacy amplification, which is also why the active phase randomisation on the CM-0015 edge of this technique repairs the proof premise without closing either route. |
CM-0014 Full Device Imperfections Incorporation in Scientific Security Proof |
|
Zhang2022 calls this its most effective countermeasure, strictly controlling the pulse period of the laser so that it cannot match the time interval between the laser and the intensity modulator. That match is what lets the response of one pulse be read from the next, since the attack light passing through one pulse shifts the centre frequency of the following one, so breaking the match removes a stated critical assumption of the isolatorless route rather than merely detecting the attack. It has no effect on the polarisation route, whose state dependence comes from Alice's polarisation optics and involves no timing relationship at all [Zhang2022]. |
CM-0041 Laser Pulse-Period Control |
|
This control is on the row because the primary defence can be silently removed. Zhang2022 states that the external isolator can be damaged by an ultrastrong light attack and its degree of isolation effectively reduced, and therefore that it must be paired with an opposite-intensity monitoring device to detect whether strong optical injection damages the external isolator. Pang2020 points the same way, noting that the attenuation of some types of attenuator can be permanently decreased after laser damage, which is the route by which Eve raises the mean photon number and with it her success rate [Zhang2022,Pang2020]. |
CM-0012 Continuous Device Functionality Monitoring |
|
State preparation monitoring checks that each emitted signal is correctly timed, appropriately polarised and emitted at the correct wavelength. Emission wavelength is precisely the parameter this technique perturbs, on both routes, so of the monitoring controls on this row it is the one whose stated scope matches the observable. Zhang2022 asks for a high-precision optical spectrum analyser against this loophole, which is the instrument this control needs to resolve a 251 MHz shift [Zhang2022]. |
CM-0016 State Preparation Monitoring |
|
Pang2020 considers placing classical detectors to monitor strong light from outside and rejects it as a sufficient defence, on the grounds that current implementations of monitoring detectors cannot be perfectly secure and have sensitivity limitations, while adding complexity and cost. At the receiver the observable is a shift in emission frequency rather than an excursion in power [Pang2020]. The control whose stated scope does match that observable is state preparation monitoring on the CM-0016 edge [Pang2020]. |
CM-0011 Watchdog Detector |
|
Component temperature is the observable that survives when forward insertion loss does not. The degradation of an isolator under high-power reverse illumination is thermal in mechanism. The absorbed light heats the magneto-optic crystal, the Verdet constant is temperature dependent, so the Faraday rotation angle drifts away from 45 degrees and the isolation falls. The signature is measurable and slow, since one circulator fell to about 35 dB and then recovers to 55 dB in 400 s, during which the sample's surface temperature decreases from 272 C to 44 C [Ponosova2022]. A temperature sensor on the first isolation component therefore sees an attack on the defence this technique depends on having succeeded, which matters because that attack is largely invisible in the forward direction [Ponosova2025]. |
CM-0040 Temperature Monitoring |
|
Optical isolation is the natural defence against injection locking and both primaries treat it as such, but neither route is closed by the isolation actually deployed. Pang2020's attack succeeded through Alice's own 30 dB built-in isolator at 60.0 percent success, and adding 1 dB and 3 dB lowered that only to 44.0 and 36.1 percent, though the paper states sufficient security is possible if Alice adds many orders more isolation. On the isolatorless route the control is not merely insufficient but architecturally excluded, because the locking light cannot pass an isolator, so an external isolator at the source output is the only place it can go [Pang2020,Zhang2022]. How much isolation is now a figure rather than a direction, and it is out of reach of isolators alone: about 140 dB of reverse attenuation [Wiesemann2024,Lovic2023]. The two child controls of this row are what close that gap, and both are linked separately on this technique, with the optical fuse bringing the requirement to about 90 dB and the optical power limiter down to approximately 30 dB [Lovic2023]. Isolation for this technique must also be counted from measured residual values behind a sacrificial first component rather than from datasheet figures, and the accounting rule is explicit, that any isolation calculated for countermeasure against the Trojan-horse attack and the laser-seeding attack shall be started from the components behind this sacrificial isolation component [Ponosova2022]. |
CM-0008 Employment of Optical Isolation |
| Description | Category | Impact Type |
|---|---|---|
|
ILA operates entirely in the optical/quantum domain: Eve injects photons into Alice's semiconductor source laser and reads the resulting frequency shift, an injection-locking effect of the laser cavity, placing it at the Physical Layer (Quantum & Optical) [Pang2020]. |
L1 — Physical Layer (Quantum & Optical) | Primary Target |
| Relation type | Description | Technique |
|---|---|---|
| Alternative to |
A-0066 Injection-Locking Attack reaches the same readout as this technique by different physics, and the pair is worth navigating between because the resemblance is easy to miss. In both, the threat actor sends polarised light backwards into the transmitter, the transmitter's own polarisation optics admit it in different amounts depending on which state is selected, and the resulting shift in the transmitter's emission frequency reveals that state. A-0066 produces the shift by locking the laser's cavity to an external reference; this technique produces it by heating the diode, which is why it works with light far outside the lasing line. The same observable, a state-dependent frequency shift in the transmitter's own emission, therefore has two distinct causes. |
A-0082 Wavelength Manipulation Attack in Free-Space |
| Combines with |
[Huang2019] section V names the injection-locking attack as a concurrent partner, in which the seeding is affected in real time by the transmitter's modulator state. The two sit in the same cat08 family and share physics: [Lovic2023] models laser seeding itself as optical injection locking using laser rate equations, reproducing the power increase, reduced turn-on delay, reduced chirp and jitter, and wavelength shift together. A-0066 is nonetheless a distinct technique, aimed at inferring the transmitter's polarisation from the frequency of Alice's emitted light rather than at raising intensity or fixing phase. |
A-0063 Laser Seeding Attack (LSA) |
|