The Bridge - February 2018 - 27

Quantum Cryptography and Side Channel Attacks

Additionally, as APDs 1-2 and APDs 3-4 both
correspond to two orthogonal basis states, we expect
a phase shift of π/2 between the polarization angles
associated with maximum measured backflash
intensities from each pair. We also expect to, with
the AD basis being the HV basis rotated by π/4, see
a phase difference of π/4 between cross-basis APD
pairs. Figure 5 shows the estimated phase values of
each slice from all four APDs, relative to each other.
This figure clearly shows four distinct phases, with
one for each APD.

Figure 5: Phase values for each primary slice from
APDs 1-4.

The data shows a clear correlation between
the backflashes from individual APDs and their
polarization. It appears that, with the appropriate
resources, a hypothetical eavesdropper would be
able exploit backflash polarization correlations to
determine the originating APD, and therefore gain
information regarding the result of the original BB84
measurement. This information leakage would not
be detected by the legitimate communicators via the
QBER, compromising QKD security.

While our quantitative results offer us confidence
in reaching our conclusion, our nonideal visibility,
distinguishability, and phase values reveal some
flaws in our apparatus. This is partially attributed to
issues such as backflash polarization distortion from
the optics used, non-ideal beamsplitters, non-ideal
circular polarization for the incoming photons from
the BB84 source, varying detector efficiencies, and
the fact that the HWP was not precisely calibrated.
In order to evaluate the practicality of the backflash
attack, one needs to quantify the rate of information
leakage due to backflashes. To do this, however,
one needs to first determine the backflash rate and
compare it to the rate of incoming photons. As one
may conclude from the discussions in [6, 4], this
value is dependent on a complicated convolution
of several functions and parameters, including the
capture efficiencies and backflash amplitudes of
the primary APDs, the efficiency for the backflash
coupling back into the quantum channel, fiber
attenuation effects, and the efficiency of the
backflash-detecting APD. Characterizing each of
those effects must be relegated to future work.
While, without the backflash rate information, we
cannot know with certainty, it may be the case
that even if an eavesdropper Eve maximizes her
detection efficiency with superior technology, the
spectral and spatial filtering countermeasures
described by Kurtsiefer et al. would reduce backflash
levels such that Eve would gain very little information
regarding the secret key. This information would then
be reduced even further with privacy amplification
procedures that the legitimate users perform to
increase the security of their key. Furthermore,
there are some emerging single-photon detection
technologies that do not use avalanches in
semiconductors, such as superconducting nanowire
detectors [8] that may not exhibit backflash behavior.

HKN.ORG

27
http://www.HKN.ORG
Table of Contents for the Digital Edition of The Bridge - February 2018
