Quantum key distribution (QKD) allows two users to communicate with theoretically provable
secrecy . This is vitally important to secure the confidential data of governments, businesses
and individuals. As the technology is adopted by a wider audience, a quantum network will
become necessary for multi-party communication, as in the classical communication networks in
use today. Unfortunately, a number of phase-encoded QKD protocols based on weak coherent
pulses have been developed. Whilst the first protocol, proposed by Bennett and Brassard
in 1984 (BB84), is still commonly used, other protocols such as differential phase shift  or
coherent one way QKD  are also adopted. Each protocol has its benefits; however all would
require a different transmitter and receiver, increasing the complexity and cost of quantum
In this work we demonstrate a multi-protocol transmitter [4-6] that also has the benefits of
small footprint, low power consumption and low complexity. We use this transmitter to give the
first experimental demonstration of an improved version of the BB84 protocol, known as the
differential quadrature phase shift protocol. We have achieved megabit per second secure key
rates at short distances, and have shown secure key rates that are, on average, 2.71 times higher
than the standard BB84 protocol. This enhanced performance over such a commonly adopted
protocol, at no expense to experimental complexity, could lead to a widespread migration to
the new protocol.
The security of the BB84 protocol relies on each signal and reference pulse pair being globally
phase randomised with respect to all other pulse pairs. In the DQPS protocol, blocks with a
length L ≥ 2 are used and each block has a globally random phase with respect to all other blocks.
Implementing this protocol would ordinarily require a high-speed random number generator and
a phase modulator. As well as increasing device complexity, it would also require an unrealistic
continuous source of electrical modulation signals for complete security. The transmitter we
use injects light from a master laser diode into a 2 GHz gain-switched slave laser diode. The
principal of optical injection locking means that the slave laser inherits the phase of the master
laser. We apply modulations to the master laser current within a block to control the phase
of the slave laser output pulses, and then drive the master laser below threshold for a short
period of time when phase randomisation is required. This ensures the lasing comes from below
threshold, thus the phase adopted by the slave laser pulse is completely random. We perform
an autocorrelation measurement on the blocks to show their randomness.
 N. Gisin et al. Rev. Mod. Phys. 74, 145 (2002).
 K. Inoue et al. Phys. Rev. Lett. 89, 037902 (2002).
 D. Stucki et al. Appl. Phys. Lett. 87 194108 (2005).
 Z. Yuan et al. Phys. Rev. X. 6, 031044 (2016).
 G. L. Roberts et al. Laser Phot. Rev. 11, 1700067 (2017).
 G. L. Roberts et al. arXiv:1709.04214 [quant-ph] (2017).
In recent years, the security of avalanche photodiodes as single photon detectors for quantum key distribution has been subjected to much scrutiny. The most prominent example of this surrounds the vulnerability of such devices to blinding under strong illumination. We focus on self-differencing avalanche photodiodes, single photon detectors that have demonstrated count rates exceeding 1 GCounts/s resulting in secure key rates over 1 MBit/s. These detectors use a passive electronic circuit to cancel any periodic signals thereby enhancing detection sensitivity. However this intrinsic feature can be exploited by adversaries to gain control of the devices using illumination of a moderate intensity. Through careful experimental examinations, we define here a set of criteria for these detectors to avoid such attacks.