independed security analysis of a commercial quantum ... · quantis from id quantique igor...
TRANSCRIPT
@ Mikhail_A_Petrov
Independed security analysis of a commercial quantumrandom number generatorQuantis from ID Quantique
Igor RadchenkoMikhail Petrov
3
Damian Steiger 4,5
Renato Renner 4
MatthiasTroyer1,6,2,7Vadim Makarov4,5
*,1,2
1
2
4
Department of Physics and Astronomy, University of Waterloo, Waterloo, ON, N2L 3G1 Canada
3
6
Russian Quantum Center, Skolkovo, Moscow 121205, Russia
NTI Center for Quantum Communications, National University of Science and Technology MISIS, Moscow 119049, RussiaMoscow State University, Moscow 119991, RussiaInstitute for Theoretical Phisics, ETH Zurich, CH-8093 Zurich, SwitzerlandMicrosoft Corporation, One Microsoft Way, Redmond, WA 98052, USAShanghai Branch, National Laboratory for Phisical Sciences at Microscale and CAS Center for Excellence in QuantumInformation, University of Science and Technology of China, Shanghai 201315, People's Republic of China
56
7
Presented at QCrypt, Amsterdam (on-line), 10-14 August 2020
YES !
NO !
Quantis Random Number Generator Broshure(from ID Quantique www.idquantique.com)
LED SLEEVE 2 Single-photondetectors
Source of quantumness
Quantis Random Number Generator from ID Quantique
We perform an independent examination of Quantis with-
out access to the manufacturer's internal documentation.
We test and analyse Quantis hardware and firmware.
What is inside ?
How it works ?
Is it really quantum ?
Is it random ?
data_sck
data_out
LED
APD2
A1
A2
DET1
DET2
QNCH NOR XOR
PP
R LD
SCK
GEN4 MHz
CNTi++INT
SSCCPLD
−Vb
APD1
∫
APD chip
Clock
“Source of quantumness”
CPLD
Linear regulators
SSC
Main printed circuit board
Simplified electrical scheme
Captured signals
flip:xn+1 = xn
hold:xn+1 = xn
S = 0 S = 1Start
A
B
flip:xn+1 = xn
hold:xn+1 = xn
A
BA⊕B
A⊕B
Measurements
Post processing algorithm
T (oC)
0 10 20 30
Dar
k co
unts
(H
z) 102
10-1
101
100
40 50 60 70
103APD's dark counts
Our full results are available in arXiv:2004.04996
We show that > 99% of Quantis output data originates in
physically random processes: random timing of photon
absorption in semiconductor material, and random growth
of avalanche owing to impact ionisation. We found minor
non random contributions from imperfecton in detector
electronics and internal processin algorithm. Howewer all
these effects stay well below specified by ID Quantique 1%.
Conclusions
We infer that statistical deviation "flip" and "hold"
probalities is due to APD efficiency mismatch (~8.8 %).
Output stream statistical test
a.
a.
b.Time interval (µs)
0.02
14
0 1 2 3
0
12
-0.02
-0.01
0.01
ln(e
vents
)
8
10
ln(e
vents
)−
linear
fit
a
b
4
18
16
-0.03
APD autocorrelation under cont. wave light
Time interval (ns)
-200 -100 0
Num
ber
of co
inci
den
ces
200
0
100
100 200 300
300
-300
APD cross-talk in darkness over 16 h
Time interval (ns)
-200 -100 0
Num
ber
of ev
ents
(x1
04)
2
0
1
100 200
3
300-300
4
APD cross-correlation under cont. wave light over 1 h