Fast gated avalanche photodiodes (APDs) are the most commonly used single photon detectors for high bit rate quantum key distribution (QKD). Their robustness against external attacks is crucial to the overall security of a QKD system or even an entire QKD network. Here, we investigate the behavior of a gigahertz-gated, self-differencing InGaAs APD under strong illumination, a tactic Eve often uses to bring detectors under her control. Our experiment and modelling reveal that the negative feedback by the photocurrent safeguards the detector from being blinded through reducing its avalanche probability and/or strengthening the capacitive response. Based on this finding, we propose a set of best-practice criteria for designing and operating fast-gated APD detectors to ensure their practical security in QKD.

The information theoretic security promised by quantum key distribution (QKD) holds as long as the assumptions in the theoretical model match the parameters in the physical implementation. The superlinear behaviour of sensitive single-photon detectors represents one such mismatch and can pave the way to powerful attacks hindering the security of QKD systems, a prominent example being the after-gate attack. A longstanding tenet is that trapped carriers causing delayed detection can help mitigate this attack, but despite intensive scrutiny, it remains largely unproven. Here we approach this problem from a physical perspective and find new evidence to support a detector's secure response. We experimentally investigate two different carrier trapping mechanisms causing delayed detection in fast-gated semiconductor avalanche photodiodes, one arising from the multiplication layer, the other from the heterojunction interface between absorption and charge layers. The release of trapped carriers increases the quantum bit error rate measured under the after-gate attack above the typical QKD security threshold, thus favouring the detector's inherent security. This represents a significant step to avert quantum hacking of QKD systems.Quantum key distribution (QKD) promises secure distribution of cryptographic digital keys [1], spurring significant development of the technology. This has rapidly matured and is now stepping out of the laboratory and into deployment in optical fibre networks [2-8]. Contributing to its maturity, a great deal of research has been devoted to quantum hacking [9][10][11][12][13], which identifies imperfections of QKD components from their theoretical models and evaluate their implications for QKD security. Best-practice criteria and countermeasures can then be developed [14-21] to reinforce the identified weak components and reclaim implementation security.Due to their exposure to the quantum channel, single photon detectors in QKD systems have been subjected to most hacking attacks in the past decade [22][23][24]. Weak detectors have been demonstrated to be under full control of an eavesdropper (Eve), resulting in a collapse of security [25]. Detector loopholes can be completely closed by novel protocols that achieve measurement-device independent security [26][27][28]. However, these protocols require an intermediate relay and therefore their deployment in the network is unfavorably complex when compared with standard point-to-point QKD links. A solution to regain detector security is thus highly desirable for relayless QKD links.Single photon detectors based on semiconductor In-GaAs avalanche photodiodes (APDs) serve the majority of links in existing QKD networks [2-7], because they operate at temperatures that are easily within reach of thermo-electric cooling [29] or even room temperature * james.dynes@crl.toshiba.co.uk [30]. The state-of-the-art systems can offer a key rate exceeding 10 Mb/s [31] and operate over 200 km fiber [32].Attacks on InGaAs APDs have revealed their vulnerabilities, most...

This is a repository copy of Backflashes from fast-gated avalanche photodiodes in quantum key distribution.