1
Demonstrating feasibility of a Trojan-horse attack on a commercial QKD system Nitin Jain , , Christoffer Wittmann , Vadim Makarov and Gerd Leuchs 1,2 1,2 3 Imran Khan , 1,2 1,2 Elena Anisimova Christoph Marquardt , 3 1,2 1 Max Planck Institute for the Science of Light, Günther-Scharowsky-Str. 1, Bau 24, 91058 Erlangen, Germany 2 Universität Erlangen-Nürnberg, Staudtstraße 7/B2, 91058, Erlangen, Germany 3 Institute for Quantum Computing, University of Waterloo, 200 University Avenue West, Waterloo, Ontario, N2L 3G1 Canada Quantum hacking: Motivation F To investigate loopholes in practical QKD that could be exploited by an eavesdropper Eve. implementations à à Imperfections & vulnerabilities mainly arise from technological deficiencies or operational weaknesses. Looking for loopholes is a normal, iterative process during development of a secure communication system. Responsible disclosure provides a positive feedback to the field. The eventual goal is to make practical implementations more secure! à N N N Quantum channel is not just for Alice and Bob, it is also a potential open door for an eavesdropper. No optical component transmits or absorbs perfectly; some light is inevitably reflected back towards the input. Optical time/frequency domain reflectometry (OTDR/OFDR) can be used to generate maps of Alice and Bob from the quantum channel, allowing Eve to know when and with what intensity to attack. N. Gisin, S. Fasel, B. Kraus, H. Zbinden, and G. Ribordy, Phys. Rev. A 73, 022320 (2006) A. Vakhitov, V. Makarov, and D. R. Hjelme, J. Mod. Opt. 48, 2023 (2001) F F F F F When exactly should the Trojan-horse pulse be launched? adversely How What Which What to choose the suitable wavelength, brightness, and polarization for this pulse? properties of the back-reflected pulse should be analyzed? components in the QKD system could be affected by the attack? percentage of the final secret key may be obtained from this attack? Eve’s bane: Afterpulsing effects in Bob’s APD based single-photon detectors à à à A back-reflection level of -60 dB, as per the OTDR map, implies Eve n millions of photons igh attack frequency worsens the problem. eeds to send (bright pulses) into Bob to get just a few photons back. Bright pulses cause in the APD due to raises the QBER and thus, may disclose Eve’s presence. afterpulsing filling of traps by charge carriers. H Afterpulsing results in substantial increase of dark counts. This R. H. Haitz, J. Appl. Phys. 36, 3123 (1965); S. Cova, A. Lacaita, and G. Ripamonti, IEEE Electron. Dev. Lett. 12 685 (1991) Clavis2 – Commercial QKD system from ID Quantique P. Jouguet et al., Phys. Rev. A 86, 032309 (2012) N. Jain, arXiv:1206.7019v1 G. Leuchs, ICQI, Ottawa (2011) V. Makarov, CLEO/Europe-EQEC, Munich (2011) V. Scarani and C. Kurtsiefer, arXiv:0906.4547v1 D. Stucki, N. Gisin, O. Guinnard, G. Ribordy, and H. Zbinden, New J. Phys. 4, 41 (2002) OTDR maps Spectral characterization of reflectivity Asymmetric double pass (both source and detector in Bob). Mach-Zehnder interferometric configuration operating in Operating principle: Plug-n-Play BB84 SARG04 and protocols with highly attenuated laser pulses ( mean photon number, < 1.0). typical μ QKD protocols implemented If Eve can surreptitiously read Bob’s modulation during the operation of , then she obtains the raw key Alice and Bob knowledge. SARG04 exchanged by without their Exploitable vulnerability State preparation, basis application and detection Alice modulates φ = {0, π/2, π, 3π/2} φ = {0, π/2} A B to prepare the state. Bob modulates to apply the basis choice and measures the outcome using avalanche photodiodes (APDs). V. Scarani, A. Acín, G. Ribordy, and N. Gisin, Phys. Rev. Lett. 92, 057901 (2004) C. H. Bennett and G. Brassard, Proc. IEEE Int. Conference on Computers, Systems, and Signal Processing, 175 (1984) A Trojan-horse eavesdropper List of questions for Eve Trojan-horse attack: from ideas to implementation Concepts Experimental results and issues for Eve quantum channel V Detector 0 Detector 1 polarizing beamsplitter 50/50 beamsplitter circulator phase modulator Bob B H Laser photodiode 90/10 coupler variable optical attenuator (VOA1) VOA2 Faraday mirror delay loops phase modulator Alice A Future directions ü ü ü Use interfaces such as with reduced reflections. Reducing the duty cycle of the phase modulation voltage. Technical countermeasures, e.g. watchdog detector, are possible but need to be adapted to security proofs. FC/APC Possible countermeasures Schematic of Alice and Bob in Clavis2 Schematic of Eve’s apparatus to perform a readout of Bob’s phase modulation Oscilloscope trace of Eve’s corresponding homodyne detection output Integrating over a suitable time window per slot for 250 slots Bob’s randomly chosen & secret basis choice slot no. in the QKD frame pulse no. in homodyne output Eve’s estimation using a suitable threshold 0 1 0 1 { { Oscilloscope trace of Bob’s phase modulation voltage 10 20 30 40 50 60 1 2 3 4 attack frequency, (kHz) f Eatt rise in dark count level 5 1 quantum bit error rate, QBER Mutual information, I(A:B) channel transmission, T fEatt =10 kHz No attack fEatt =20 kHz 0.1 0.2 0.3 0.4 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.2 0.4 0.6 0.8 } ΔQBER = +2.3% } ΔI(A:B) = -8.8% Increase in dark counts due to afterpulsing and corresponding changes in quantum bit error rate, QBER [solid lines] and mutual information of Alice and Bob, I(A:B) [dashed lines]. Above 99% correlations! C. Branciard, N. Gisin, B. Kraus, V. Scarani, Phys. Rev. A 72, 032301 (2005) D G N M A We are devising an attack Preliminary simulation results show this strategy works fterpulsing is a blessing in disguise for security. strategy in which Eve manipulates the QKD frame sent by Alice to Bob so that: for low channel transmission values ( 0.2) at least. Spectral sensitivity of the APD is four orders smaller at 1700 nm than at 1550 nm. We conjecture the afterpulsing might be significantly reduced in this wavelength regime. T Attack strategy and further improvements à à the incurred QBER is below the abort-threshold of Clavis2, the privacy amplification step is insufficient in preventing Eve from knowing a partial amount of the final secret key, and the detection rate observed by Bob is within the bounds of the expected detection rate. à Control Arm 50:50 BS Signal Arm Homodyne detector Outcoupler FC1 on a micrometer stage fiber coupler Eve laser synchronized to Bob isolator Bob 1 Bob PC1 PC3 4 3 9 2 coarse delay line PC2 FC2 PC: polarization controller BS: beam splitter FC: Fiber collimator Results of phase readout; mean photon number, 100\10 and attack frequency, = 5 MHz μ f 8 Eatt (Signal\Control arm) φ π Bob = /2 φ Bob = 0 0.2 0.4 0.6 0.8 1.0 time (μs) time (μs) Reflections (in dB) time delay (ns) 1550 nm Sum of the remaining three connectors 806 nm D0 Laser D1 PM PBS-BS-C D0 Laser D1 D0 Laser D1 short-short short-long / long-short long-long PBS-BS-C PBS-BS-C -40 -100 -60 -120 -80 -40 -100 -60 -120 -80 0 20 40 60 80 100 120 140 No. of publications Year Inception of MagiQ Technologies and ID Quantique 2000 2002 2004 2006 2008 2010 2012 1 2 3 4 5 6 SeQuereNet is launched -70 -60 -50 -40 -30 -20 -10 Wavelength (nm) Supercontinuum OSA Noise Floor 800 1000 1200 1400 1600 power (dBm) Reflection from Bob with connector after PM open Reflection from Bob as a whole

Demonstrating feasibility of a Trojan-horse attack on a … · 2017. 3. 22. · Demonstrating feasibility of a Trojan-horse attack on a commercial QKD system Nitin Jain , , Christoffer

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Demonstrating feasibility of a Trojan-horse attack on a … · 2017. 3. 22. · Demonstrating feasibility of a Trojan-horse attack on a commercial QKD system Nitin Jain , , Christoffer

Demonstrating feasibility of a Trojan-horse attack on a commercial QKD systemNitin Jain , , Christoffer Wittmann ,Vadim Makarov and Gerd Leuchs

1,2 1,2

3

Imran Khan ,1,2

1,2

Elena Anisimova Christoph Marquardt ,3 1,2

1Max Planck Institute for the Science of Light, Günther-Scharowsky-Str. 1, Bau 24, 91058 Erlangen, Germany

2Universität Erlangen-Nürnberg, Staudtstraße 7/B2, 91058, Erlangen, Germany

3Institute for Quantum Computing, University of Waterloo, 200 University Avenue West, Waterloo, Ontario, N2L 3G1 Canada

Quantum hacking: Motivation� To investigate loopholes in practical QKD that

could be exploited by an eavesdropper Eve.implementations

Imperfections & vulnerabilities mainly arise from technologicaldeficiencies or operational weaknesses.

Looking for loopholes is a normal, iterative process duringdevelopment of a secure communication system.

Responsible disclosure provides a positive feedback to the field.The eventual goal is to make practical implementations more secure!�

Quantum channel is not just for Alice and Bob, it is also a potential open door for an eavesdropper.

No optical component transmits or absorbs perfectly; some light is inevitably reflected back towards the input.

Optical time/frequency domain reflectometry (OTDR/OFDR) can be used to generate maps of Alice and Bob fromthe quantum channel, allowing Eve to know when and with what intensity to attack.

N. Gisin, S. Fasel, B. Kraus, H. Zbinden, and G. Ribordy, Phys. Rev. A 73, 022320 (2006)A. Vakhitov, V. Makarov, and D. R. Hjelme, J. Mod. Opt. 48, 2023 (2001)

When exactly should the Trojan-horsepulse be launched?

adversely

How

What

Which

What

to choose the suitable wavelength,brightness, and polarization for this pulse?

properties of the back-reflectedpulse should be analyzed?

components in the QKD systemcould be affected by the attack?

percentage of the final secret keymay be obtained from this attack?

Eve’s bane: Afterpulsing effects in Bob’s APD based single-photon detectors

A back-reflection level of -60 dB, as perthe OTDR map, implies Eve nmillions of photons

igh attack frequency worsens the problem.

eeds to send(bright pulses) into Bob

to get just a few photons back.

Bright pulses cause in theAPD due to

raises the QBER andthus, may disclose Eve’s presence.

afterpulsingfilling of traps by charge carriers.

H

Afterpulsing results in substantial increaseof dark counts. This

R. H. Haitz, J. Appl. Phys. 36, 3123 (1965); S. Cova, A. Lacaita, and G. Ripamonti, IEEE Electron. Dev. Lett. 12 685 (1991)

Clavis2 – Commercial QKD system from ID Quantique

P. Jouguet et al., Phys. Rev. A 86, 032309 (2012)N. Jain, arXiv:1206.7019v1

G. Leuchs, ICQI, Ottawa (2011)V. Makarov, CLEO/Europe-EQEC, Munich (2011)

V. Scarani and C. Kurtsiefer, arXiv:0906.4547v1D. Stucki, N. Gisin, O. Guinnard, G. Ribordy,

and H. Zbinden, New J. Phys. 4, 41 (2002)

OTDR maps

Spectral characterization of reflectivity

Asymmetricdouble pass (both source and detector in Bob).Mach-Zehnder interferometric configuration

operating in

Operating principle: Plug-n-Play

BB84 SARG04and protocols with highly attenuatedlaser pulses ( mean photon number, < 1.0).typical μ

QKD protocols implemented

If Eve can surreptitiously read Bob’s modulation during theoperation of , then she obtains the raw key

Alice and Bob knowledge.SARG04 exchanged

by without their

Exploitable vulnerability

State preparation, basis application and detection

Alice modulates φ = {0, π/2, π, 3π/2}

φ = {0, π/2}A

B

to prepare the state.

Bob modulates to apply the basis choice and

measures the outcome using avalanche photodiodes (APDs).

V. Scarani, A. Acín, G. Ribordy, and N. Gisin, Phys. Rev.Lett. 92, 057901 (2004)

C. H. Bennett and G. Brassard, Proc. IEEE Int. Conference onComputers, Systems, and Signal Processing, 175 (1984)

A Trojan-horse eavesdropper

List of questions for Eve

Trojan-horse attack: from ideas to implementationConcepts

Experimental results and issues for Eve

quantumchannel

V

Detector 0 Detector 1

polarizingbeamsplitter

50/50 beamsplitter

circulator

phase modulator

BobB

H

Laser

photodiode

90/10coupler

variableopticalattenuator(VOA1)

VOA2

Faradaymirror

delay loops

phase modulator

Alice

A

Future directions

Use interfaces such as with reduced reflections.

Reducing the duty cycle of the phase modulation voltage.

Technical countermeasures, e.g. watchdog detector, arepossible but need to be adapted to security proofs.

FC/APC

Possible countermeasures

Schematic of Alice and Bob in Clavis2

Schematic of Eve’s apparatus to perform a readout of Bob’s phase modulation

Oscilloscopetrace of Eve’scorresponding

homodynedetection output

Integratingover a

suitabletime windowper slot for250 slots

Bob’s randomlychosen & secret

basis choice

slot no. in the QKD frame

pulse no. in homodyne output

Eve’sestimation

using asuitable

threshold

0

1

0

1

{{

Oscilloscopetrace of Bob’s

phase modulationvoltage

10 20 30 40 50 60

1

2

3

4

attack frequency, (kHz)fEatt

rise

in d

ark c

ount

level 5

1 qu

antu

m b

it e

rro

r ra

te,Q

BE

R Mu

tual in

form

ation

,I(A

:B)

channel transmission, T

fEatt =10 kHzNo attack

fEatt =20 kHz

0.1 0.2 0.3 0.4

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.2

0.4

0.6

0.8

}

ΔQBER = +2.3%

}

ΔI(A:B) = -8.8%

Increase in dark counts due to afterpulsing and correspondingchanges in quantum bit error rate, QBER [solid lines] and mutualinformation of Alice and Bob, I(A:B) [dashed lines].

Above 99%

correlations!

C. Branciard, N. Gisin, B. Kraus, V. Scarani, Phys. Rev. A 72, 032301 (2005)

A

We are devising an attack

Preliminary simulation results show this strategy works

fterpulsing is a blessing in disguise for security.

strategy in which Evemanipulates the QKD frame sent by Alice to Bob so that:

for low channel transmission values ( 0.2) at least.

Spectral sensitivity of the APD is four orders smaller at1700 nm than at 1550 nm. We conjecture the afterpulsingmight be significantly reduced in this wavelength regime.

T ≤

Attack strategy and further improvements

the incurred QBER is below the abort-threshold of Clavis2,

the privacy amplification step is insufficient in preventingEve from knowing a partial amount of the final secret key, and

the detection rate observed by Bob is within the bounds ofthe expected detection rate.�

Control Arm

50:50 BS

Signal Arm

Homodynedetector

Outcoupler FC1 ona micrometer stage

fibercoupler

Eve lasersynchronizedto Bob

isolator

Bob

1

Bob

PC1

PC3

4

3

9

2

coarse delayline

PC2

FC2

PC: polarization controllerBS: beam splitterFC: Fiber collimator

Results of phase readout; mean photon number, 100\10 and attack frequency, = 5 MHzµ f≈

8

Eatt(Signal\Control arm)

φ πBob = /2

φBob = 0

0.2 0.4 0.6 0.8 1.0time (μs)

time (μs)

Reflections (

in d

B)

time delay (ns)

1550 nm

Sum of the remainingthree connectors 806 nm

D0Laser

D1

PM

PBS-BS-CD0

LaserD1

D0Laser

D1

short-short short-long / long-short long-long

PBS-BS-C

PBS-BS-C

-40

-100

-60

-120

-80

-40

-100

-60

-120

-80

0 20 40 60 80 100 120 140

No

. o

f p

ub

lica

tio

ns

Year

Inception ofMagiQ Technologiesand ID Quantique

2000 2002 2004 2006 2008 2010 2012

123456

SeQuereNetis launched

-70

-60

-50

-40

-30

-20

-10

Wavelength (nm)

Supercontinuum

OSA Noise Floor

800 1000 1200 1400 1600

po

we

r (d

Bm

)

Reflectionfrom Bob with

connector afterPM open

Reflectionfrom Bob

as a whole