Passive Faraday Mirror Attack
The Passive Faraday Mirror Attack targets phase-encoded QKD systems by exploiting imperfections in Faraday mirrors (FMs), which fail to provide perfect polarization rotation. A Faraday mirror is meant to return light with its polarisation turned by exactly the right amount, and a real one is slightly off. That small error makes the transmitter (Alice) stamp its encoding choice onto the polarisation of the light as well as onto the property it meant to use, which allows the threat actor (Eve) to partially discern the polarization state of the quantum signals. On the two-way plug-and-play route she replaces the receiver (Bob) on the line, measures what the transmitter returns, sends ordinary states on to him, and thereby gains partial knowledge of the cryptographic key while causing fewer errors than a plain interception would [Sun2011].
How much this matters depends on the architecture, and the two cases differ sharply. In a two-way plug-and-play system, where the light makes a round trip through the transmitter's equipment, the attack is a live concern and gets cheaper still when run alongside the phase-remapping attack. In a one-way Faraday-Michelson system the same defect leaks, but the architecture fixes a setting the threat actor needs to vary, so the interception itself is not realizable there and what remains is a reduction of the secure key rate [Wang2013].
Two practical limits bound the whole technique. It works only on a long, lossy fibre link, because the threat actor succeeds on very few pulses and has to hide the rest inside loss the receiver already expects [Sun2011]. Neither paper ran it on hardware: both are analysis and simulation.
| [ETSI011] | ETSI. "Quantum Key Distribution (QKD); Component characterization: characterizing optical components for QKD systems". (2016) https://www.etsi.org/deliver/etsi_gs/QKD/001_099/011/01.01.01_60/gs_QKD011v010101p.pdf. |
| [Qi2021] | R. Qi, H. Zhang, J. Gao, L. Yin, and G.-L. Long. "Loophole-free plug-and-play quantum key distribution" (2021) 10.1088/1367-2630/ac0478. arXiv:2010.05503. |
| [Sun2011] | S. Sun, M. Jiang, and L. Liang. "Passive Faraday-mirror attack in a practical two-way quantum-key-distribution system" In: Phys. Rev. A 83, 062331. (2011) 10.1103/PhysRevA.83.062331. |
| [Wang2013] | W. Wang, M. Gao, and Z. Ma. "Effect of imperfect Faraday mirrors on the security of a Faraday-Michelson quantum cryptography system" In: J. Phys. A: Math. Theor. 46 455301. (2013) 10.1088/1751-8113/46/45/455301. |
4440d5aa-2522-4bc9-a68f-bb2e35e7dc50
- Quantum Communication
Technique properties
Operational properties
| Description | Countermeasure |
|---|---|
|
The answer Wang2013 itself gives, and it is a proof-side answer rather than a repair: the imperfection stays in the device and the security analysis absorbs what it leaks. Following the security proof of Marøy, Lydersen and Skaar for arbitrary individual device imperfections, the paper derives a modified infinite-length key rate whose basis-dependence term is a fidelity that characterizes the basis dependence of the source and is a function of the mirror rotation errors, with the transmitter's encoded-arm error dominating and the receiver's contributing a smaller penalty [Wang2013]. Its closing advice is addressed to operators in those terms, that the security analysis can tell legal parties how to adjust privacy amplification to keep a lookout for a potential eavesdropper [Wang2013]. The leaked information is then removed in privacy amplification at the cost of key length, which is why this is recorded as neutralising the attack rather than preventing it. |
CM-0014 Full Device Imperfections Incorporation in Scientific Security Proof |
|
Fitting and aligning the Faraday mirror closer to its ideal angle, which is a device-design measure against a device-design defect. The lever is unusually sharp. At exactly the ideal angle the four states the transmitter returns fall back to two dimensions and the attack is invalid in that point [Sun2011], so the defect has a true zero rather than an asymptote. Short of zero, tightening the tolerance removes the threat actor's opportunity rather than just reducing her yield: at a one-degree error she needs a link lossier than 2.43e-3, which corresponds to a 124 km long fiber, and below 0.65 degrees she needs more than 142.5 km, because her success probability depends on the degree of the imperfection rapidly while the error rate she induces barely moves with it [Sun2011]. |
CM-0030 Quantum Device Optimal Design and TEMPEST |
| Description | Category | Impact Type |
|---|---|---|
|
Post-Processing and Cryptographic, as a secondary effect, because that is where the attack is paid for and where the one-way route lands entirely. Whether the attack succeeds is decided against a post-processing parameter: 14.64 per cent is below the 20 per cent tolerated under two-way post-processing but above the 11 per cent collective-attack bound for one-way post-processing, and the combined variant at 3.57 per cent crosses below both [Sun2011]. On the one-way route, where the interception is judged neither optimal nor realizable, what remains is a modified secret-key rate and advice on how to adjust privacy amplification to keep a lookout for a potential eavesdropper [Wang2013]. A system that does not adjust it keeps distributing a key its proof no longer covers. Moderate confidence, since neither paper uses layer language and the mapping is read from where the thresholds and the remedy sit. |
L5 — Post-Processing & Cryptographic | Secondary Effect |
|
Physical Layer, Quantum and Optical, is the primary target. The attack acts entirely in the optical domain: it exploits the angular setting of a passive optical component, so that the information encoded by Alice appears not only in the time mode but also in the polarization mode and the four returned states span three optical dimensions instead of two, and the threat actor separates them with a five-element POVM in three dimensions before resending a standard state [Sun2011]. Nothing electronic, firmware-level or protocol-level has to be wrong for the attack to work. |
L1 — Physical Layer (Quantum & Optical) | Primary Target |
| Relation type | Description | Technique |
|---|---|---|
| Combines with |
The two attacks are explicitly combined by this row's own primary, and the combination is what makes this one dangerous rather than merely detectable. Alone, the passive Faraday-mirror attack induces 14.64 per cent, and Sun2011 states that the QBER induced by Eve's attack is much lower than the 25 per cent of a general intercept-and-resend attack, though it is still above the 11 per cent collective-attack bound. Running the phase-remapping attack alongside it lets the inter-state phase leave pi/2 and range over the interval, which takes the induced error as low as about 3.6 per cent, below 11 per cent [Sun2011]. The two contribute different things: phase remapping has a much larger success probability but a higher error rate on its own, 17.7 per cent against this attack's 4.71 per cent at the same phase, so each covers the other's weakness [Sun2011]. The combination is unavailable in the one-way Faraday-Michelson system, where the modulation step is pinned [Wang2013]. |
A-0054 Phase-Remapping Attack |
| Refines |
This technique is an intercept-and-resend attack with the error cost bought down by a device defect, which is how both primaries describe it. Wang2013 states it plainly, that the attack is just an individual attack based on intercept-resend attack whose minimal error rate of about 14.64 per cent is just the tolerable upper bound of error rate in individual attack, so that it is neither optimal nor realizable in this QKD system [Wang2013]. The refinement is quantitative and specific: a plain intercept-and-resend on BB84 costs about 25 per cent because the threat actor must guess a basis, while here the rotation error lifts the four states into three dimensions so she discriminates them better, and Sun2011 states that the QBER induced by Eve's attack is much lower than 25 per cent, at 14.64 per cent or 3.57 per cent with phase remapping [Sun2011]. A-0045's own body records that intercept-and-resend is used as a subaction within more complex quantum attacks, which is the role it plays here. |
A-0045 Intercept-and-Resend Attack |
| Relation type | Description | Technique |
|---|---|---|
| Combines with |
The second composition the paper names, gated on the transmitter's Faraday mirror being imperfect. Where it is, the threat actor obtains more information by layering the passive Faraday-mirror attack on top of this one, since both target the same two-way return path [Sun2012]. |
A-0056 Partially-Random-Phase Attack |
|