See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224453455

Differential phase shift-quantum key distribution

Article in IEEE Communications Magazine · June 2009

DOI: 10.1109/MCOM.2009.4939284 · Source: IEEE Xplore

CITATIONS READS

5 2,625

4 authors, including:

Toshimori Honjo Yasuhiro Tokura

Nippon Telegraph and Telephone University of Tsukuba
114 PUBLICATIONS 5,463 CITATIONS 317 PUBLICATIONS 10,105 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Yasuhiro Tokura on 26 November 2015.

The user has requested enhancement of the downloaded file.

 Selected Papers

Differential-phase-shift Quantum Key

Distribution
Toshimori Honjo† and Kyo Inoue
Abstract
Quantum cryptography promises to provide communications with unconditional security on the basis
of the physical principles of quantum mechanics. We have proposed a quantum key distribution proto-
col called a differential-phase-shift quantum key distribution (DPS-QKD). This scheme has several
advantages including suitability for fiber transmission and highly efficient key generation. In this paper,
we review quantum cryptography technologies and then describe our scheme DPS-QKD. In an experi-
ment using a planar lightwave circuit Mach-Zehnder interferometer, DPS-QKD achieved stable polar-
ization-independent operation during 20-km fiber transmission. The raw key creation rate was 3076 bit/s
with a 5.0% quantum bit error rate.

1. Introduction and background power but on the laws of quantum mechanics [1].

Secure communications is one of the most impor- 1.1 Quantum key distribution
tant topics in the telecommunication field. At present, Quantum cryptography allows two physically sep-
cryptography, the art of hiding information in a string arated parties to create a random secret key. For this
of bits meaningless to any unauthorized party, is usu- reason, it is often called quantum key distribution
ally used for secure communications. (QKD). Once a sender (Alice) and a receiver (Bob)
Classical cryptography, whose security is based on share random secret keys, they can achieve perfectly
the need for a large amount of computation, is wide- secure communication using a one-time-pad method,
ly used in today’s communications systems. Howev- which has been proved to provide complete security.
er, this type of cryptography will be vulnerable to In short, QKD is a technique for establishing a
attack from clever eavesdroppers in the future. For quantum channel on which Alice and Bob can detect
example, RSA public key cryptography (named after the existence of an eavesdropper. When this quantum
Rivest, Shamir, and Adleman) is based on the diffi- channel is available, Alice and Bob can create a key
culty of factoring large integers. If an efficient way to through it and can be assured of the security of the
factorize large integers is found, RSA will no longer key. The quantum channel is constructed based on the
be secure. In fact, with the advent of quantum com- following theorem in quantum mechanics: an eaves-
puters, it will be possible to factor large numbers in dropper (Eve) cannot acquire any information from a
an instant. Therefore, a lot of activity is being focused quantum state transmitted from Alice to Bob without
on finding cryptography schemes that will provide disturbing its quantum state, so when Eve acquires
ultimately secure communications. some information from the quantum state, the state is
Now quantum cryptography is in the spotlight. perturbed and a key bit error is induced. This theorem
Even if a quantum computer is made, quantum cryp- is deduced from the following laws of quantum
tography provides ultimately secure communications mechanics.
because its security is based not on computational (1) With an appropriate measurement basis, a cer-
tain physical property of a quantum state is determi-
† NTT Basic Research Laboratories nately measured. But with another measurement
Atsugi-shi, 243-0198 Japan basis, the physical property is measured probabilisti-
E-mail: [email protected] cally.

26 NTT Technical Review

 Selected Papers

(2) No one can clone an unknown quantum state in error-corrected keys using a universal hash function.
general. Then they distill a fully secret key, which can be used
Based on the above idea, Alice and Bob can gener- with full confidence to encrypt a message.
ate a secret key as follows. Alice encodes the bit
information onto quantum states and sends it to Bob 1.2 Conventional QKD protocol and
through the quantum channel. Bob randomly chooses implementation
a measurement basis for each quantum state and mea- Several protocols have been proposed for the above
sures the state. After that, Bob discloses the measure- idea, such as BB84 [2], B92 [3], and E91 [4]. Here,
ment basis and then Alice and Bob choose the proper we overview BB84, the most famous QKD protocol,
bits, which could be a raw key. To check the existence to explain the above idea. We also introduce Plug &
of an eavesdropper, Alice and Bob choose a fraction Play QKD, which is a widely performed fiber-based
of their raw keys and compare them over a public BB84-QKD implementation.
(classical) channel. If an eavesdropper measures the 1.2.1 BB84 protocol
quantum state to steal the information, the quantum BB84, which was proposed by Bennett and Bras-
state changes and a key bit error is induced with some sard in 1984, is a QKD protocol using four quantum
probability. Thus, from the error rate in the test bits, states with two non-orthogonal bases. In practice, a
Alice and Bob can detect whether an eavesdropper single photon is used as a quantum state. In imple-
exists or not and also evaluate the amount of leaked menting the BB84 protocol, there are two ways to
information, assuming that all errors, including those encode bit information onto a single photon: polar-
due to imperfections in practical systems, result from ization encoding and phase encoding. Here, we
eavesdropping. If the estimated information leakage describe polarization encoding because it is easy to
exceeds an upper bound, they discard their raw keys. understand intuitively.
If the error rate or leakage is small, they perform the Figure 1 shows the sequence of the BB84 protocol.
following sequence. First, Alice and Bob perform First, Alice sends single photons in one of four polar-
error correction to obtain a matched bit string. The ization states: vertical linear, horizontal linear, right
obtained string may not be completely private. To circular, or left circular. She randomly chooses one of
eliminate the possibility of leakage, Alice and Bob the polarization states for each photon and records
perform privacy amplification [1], which shortens the her choice. Bob has two filters and selects one of

Eve
Filter 1

Filter 2

Eavesdropping
Intercept Resend
Filter 1
Alice Bob
Polarization Right Horiz. Left Vert.
state
Filter 2

Photon Quantum channel

Sent data Measurement results

Classical channel
Bob’s measurement biases Measurement biases

Bias-matched data Bias-matched data

Raw key Raw key

0 1 0 0 1 0

Fig. 1. Sequence of the BB84 protocol using polarization encoding.

Vol. 2 No. 12 Dec. 2004 27

Selected Papers

them randomly before measuring each photon. One posed by Muller et al. [5]. In this system, instead of
of the filters allows him to distinguish between hori- being encoded onto polarization states, the bit infor-
zontally and vertically polarized photons, whereas mation is encoded onto the relative phase of the
the other distinguishes between right and left circu- superposition of single photon states. This is what is
larly polarized ones. When Bob chooses the matched known as phase coding. Most experiments and com-
filter, he gets the correct output. When Bob chooses mercial products use this QKD setup.
the filter that does not correspond to the polarization Figure 2 shows the schematic diagram of the Plug
state sent by Alice, each outcome can occur with 50% & Play QKD system. Bob injects a light pulse
probability. Bob records the filters used and the out- through a circulator. The pulse is split into two puls-
comes. es at a coupler. One pulse takes the short arm and the
Second, after receiving a sufficient number of pho- other takes the long arm. A polarization controller is
tons, Bob announces through a public channel which set in each arm so that the pulse is completely trans-
filter he used for each photon, but does not reveal the mitted at the polarization beam splitter. The pulses
measurement results. Alice compares Bob’s data with propagate to Alice and are reflected by a Faraday mir-
the list of states she sent, and tells Bob for which pho- ror. One pulse is phase modulated at PM(a) by φa = {0
or π or π or 3π } at Alice’s site. The light power is
tons he used a compatible filter, but not the polariza-
tion states themselves. In cases where states and fil- 2 2
ters are compatible, they keep the data. Otherwise the attenuated to be 0.1 photon per pulse, and then the
data are simply discarded. Under the agreement that pulses travel back to Bob. Thanks to the effect of the
vertical linear or right circular polarization states Faraday mirror, the birefringence of the optical fiber
denote a “0” bit and horizontal linear or left circular is automatically compensated, and the pulse comes
polarization states denote a “1” bit, Alice and Bob back orthogonally polarized. Each pulse travels
can share the same bit string, which could be a raw through the counter arm. The light that passes
key. through the long arm is phase modulated at PM(b) by
φb = {0 or π } at Bob’s site. Since the pulses travel
Third, Alice and Bob each choose a fraction of their
raw keys and compare them over a public channel to 2
assess the secrecy of their communication. They per- through the same optical path, they stably interfere
form error correction and privacy amplification, as with each other at the coupler. The relative phase
described in the previous section. Then, Alice and between φa at Alice’s site and φb at Bob’s determines
Bob can share a secret key, which can be used with which photon detector clicks, DET1 or DET2.
perfect confidence to encrypt a message. With this setup, Alice randomly chooses one of four
phase modulations: 0, π , π, or 3π , which corresponds
1.2.2 Plug & Play QKD
Though QKD using polarization states is easy to 2 2
understand, it is not suitable for fiber transmission, to the choice of polarization states mentioned in the
because the polarization state is not maintained BB84 protocol explanation. Bob randomly chooses one
of two phase modulations: 0 or π , which corresponds
because of the birefringence in optical fiber. A QKD
setup suitable for fiber transmission has been pro- 2

Bob

Alice
Polarization DET1
Faraday Phase beam
Fiber splitter
mirror modulator Attenuator DET2
PM(a)
Phase 50:50
π 3π Circulator
φa = 0, , π, modulator coupler
2 2
PM(b)
Intensity
π modulator
φb = 0,
2
Laser diode

Fig. 2. Schematic diagram of the Plug & Play QKD system.

28 NTT Technical Review

 Selected Papers

to the choice of filters. 2. Differential-phase-shift QKD

After receiving a sufficient number of photons, Bob
announces on a public channel which phase modula- We proposed DPS-QKD [6] to overcome the disad-
tion he imposed on each pulse. vantages of conventional QKD protocols. This
Alice compares this sequence with the list of her scheme uses the uncertainty of the detection time of
phase modulations and tells Bob for which photons photons and encodes the bit information onto the
he imposed compatible phase modulations. phase difference between two sequential pulses.
Under the agreement that the phase difference φa –
φb of 0 denotes a “0” bit and the phase difference φa – 2.1 Protocol
φb of π denotes a “1” bit, Alice and Bob can share the Figure 3 shows the setup and protocol sequence of
same bit string, which could be a raw key. the DPS-QKD scheme. Alice randomly phase-modu-
The great advantage of this setup is that the polar- lates a pulse train of weak coherent states by {0, π}
ization change in the transmission line is automati- for each pulse and sends it to Bob with an average
cally compensated. However, this scheme has some photon number of less than one per pulse. Bob
disadvantages. First, the light traveling in an optical divides each incoming pulse into two paths and
fiber is scattered by inhomogeneities. A small frac- recombines them with a 50:50 beam splitter, where
tion of the light is recaptured by the fiber in the back- the path-length difference is set equal to the time
ward direction, which is called Rayleigh backscatter- interval of the sequential pulses. Photon detectors are
ing. Because of the intrinsically bi-directional nature placed at the two outputs of the recombining beam
of the system, backscattered photons induce errors. splitter. At the detectors, the partial wave functions of
To avoid this Rayleigh backscattering, the repetition two sequential pulses interfere with each other, as
frequency must not be too high. illustrated in Fig. 3. With an appropriate phase in the
Second, light transmitted through the fiber must be interferometer, detector 1 clicks for 0 phase differ-
phase-modulated at Alice’s site, so polarization con- ence between the two consecutive pulses and detector
trol is needed. 2 clicks for π phase difference.
Third, there are many components, such as a phase- Using the above setup, Alice and Bob create a raw
modulator, polarization controllers, a polarization key by the following protocol. After raw transmis-
beam-splitter, and a circulator, at Bob’s site. These sion, Bob tells Alice the time instances at which a
components increase the excess loss, so many of the photon was counted. From this time information and
photons that have traveled back from Alice vanish at her modulation data, Alice knows which detector
Bob’s site. clicked at Bob’s site. Under the agreement that a click

Bit information Bob Phase

Setup 0 1 1 0
1-bit delay DET1 difference
Alice π π 0 π π circuit 0
< 1 photon
Coherent per pulse DET2
light π
source Phase Attenuator
π π 0 π π
modulator
π π 0 π π
π π 0 π π

t4 t3 t2 t1 Time
Protocol
Alice Bob
Time t1 t2 t3 t4 t5 t6 t7 Time t2 t4 t6
Phase Detector Det2 Det1 Det2
0 π 0 0 0 π 0
difference
Phase
π 0 π
Time t2 t4 t6 difference

Raw key Raw key

1 0 1 1 0 1
bits bits

Fig. 3. Setup and protocol of the DPS-QKD scheme.

Vol. 2 No. 12 Dec. 2004 29

Selected Papers

by detector 1 denotes “0” and a click by detector 2 Bob’s long path. In the second case the detector clicks
denotes “1”, for example, Alice and Bob obtain an according to the phase difference between the two
identical bit string. time slots, which gives a correct answer. Bob does not
notice any eavesdropping in this case. However, no
2.2 Eavesdropping interference occurs and the detectors click randomly
DPS-QKD uses the uncertainty of the detection at the first and third time instances. A bit error is
time of photons. If an eavesdropper steals a photon introduced by these detection events. The probability
and sends a fake one, Alice and Bob notice the exis-
of clicks at the first or third time instances is 1 , so
tence of the eavesdropper from the bit error induced 2
by the detection timing error. 1
the error rate is . The eavesdropping is revealed
Here, we show one example of security against one 4
type of intercept-resend attack. Eve intercepts and from this error rate.
measures transmitted states and resends a false signal
to Bob according to her measurement, as illustrated 2.3 Features
in Fig. 4. Eve cannot measure every phase difference DPS-QKD has several advantages over convention-
because the transmitted state is less than one photon al QKD schemes.
per pulse. Eve thus sends a signal only when she First, this scheme is polarization insensitive, pro-
detects a photon. She sends a single photon split into vided that a polarization insensitive interferometer is
two time slots through an interferometer identical to available. In our scheme, the information is carried by
Bob’s, in which the relative phase between the two the phase difference between two sequential pulses.
time slots is 0 or π depending on the measured phase Though the polarization state changes after propaga-
difference. For unmeasured time slots, on the other tion through the fiber, two sequential pulses experi-
hand, she sends no photon. This fake signal generates ence the same change so they have the same polar-
the same count rate in Bob’s detectors as the original ization state at the fiber output, as long as the time
one. Bob does not notice the eavesdropping from the interval of two sequential pulses is much shorter than
photon counting rate. However, a bit error is intro- the time constant of the change in the fiber. This con-
duced from this fake signal as follows. When a pho- dition is satisfied in actual systems because changes
ton split into two time slots arrives at Bob’s site, he in temperature and/or mechanical pressure have a
could count a photon at one of three possible time slow time constant compared with the pulse interval.
instances: (1) when a photon passes through the short Therefore, nearly perfect interference between two
path in Eve’s interferometer and the short path in sequential pulses is possible independent of polariza-
Bob’s interferometer, (2) when a photon passes tion change in fiber.
through Eve’s short path and Bob’s long path or Second, DPS-QKD has a simpler receiver setup and
through Eve’s long path and Bob’s short path, or (3) thus smaller loss than other QKD setups. For exam-
when a photon passes through Eve’s long path and ple, in the Plug & Play QKD setup, there are (as men-

Eve
1-bit delay DET1
circuit
DET2
π π
0

Bob
1-bit delay DET1
Alice circuit
DET2

t3 t2 t1
Time

Fig. 4. Intercept-and-resend attack.

30 NTT Technical Review

 Selected Papers

tioned before) many components, which increase the 2.4.1 Performance of the PLC Mach-Zehnder
excess loss at the receiver’s site. In DPS-QKD on the interferometer
other hand, there is only one interferometer at the Before conducting the transmission experiment, we
receiver’s site so the excess loss is much smaller. Very evaluated the performance of the PLC Mach-Zehnder
few photons are lost at the receiver’s site, which interferometer. We used one packaged with input and
means we can achieve a high key generation rate. output fibers. A peltier device and a thermistor were
Third, key generation efficiency is high compared attached to the PLC chip to control the temperature,
with conventional QKD protocols. In the BB84 pro- so the phase difference between the two waveguide
tocol, half of the received photons are discarded paths could be stably controlled. The path-length dif-
because they are basis-mismatched. On the other ference was 20 cm, which introduced a one-bit delay
hand, the DPS-QKD protocol utilizes all photons for of 1 ns at 1 Gbit/s. The excess loss of the interferom-
creating a key, which provides high efficiency. eter was 2.64 dB (fiber-to-fiber). The PLC is sensitive
Fourth, in the DPS-QKD setup, it is possible to to the polarization state due to its birefringence
send a light pulse at a high repetition frequency. In the (about 10–4), which is induced by the residual stress
Plug & Play QKD setup, the repetition frequency in the silica glass film. This results in a polarization-
cannot be made too high because of Rayleigh dependent spectral response shift and degrades the
backscattering. On the other hand, DPS-QKD is a extinction ratio. The shift can be eliminated using a
one-way transmission system so the light can be birefringence compensation technique or by choos-
transmitted at a high repetition frequency and we can ing the wavelength at which the shift is zero. To eval-
achieve a high key generation rate. uate the stability of the interferometer, we measured
the extinction ratio at various temperatures for the
2.4 Experiment best and worst polarization states. Figure 5 shows the
As described above, the DPS-QKD scheme has results, where the upper and lower lines are the
some advantages over conventional QKD schemes. extinction ratios for the worst and best polarization
One of the most characteristic advantages is the states, respectively. Even for the worst polarization
polarization insensitive operation. However, this pre- state, an extinction ratio of less than –20 dB (1%) was
supposes that a stable and polarization insensitive obtained under temperature control within 0.05°C. At
interferometer is available. We performed an experi- the optimum temperature, the polarization depen-
ment using a Mach-Zehnder interferometer fabricat- dence was small, so that the extinction ratio ranged
ed using a planar lightwave circuit (PLC) based on from 0.27 to 0.46% when the input polarization state
silica waveguide technologies [7]-[9]. was varied. These results show that stable polariza-
tion-independent operation is possible using a PLC

–10
Extinction ratio (dB)

–15

–20

–25

21.7 21.74 21.78 21.82 21.86

Temperature (°C)

Fig. 5. Extinction ratio of the PLC Mach-Zehnder interferometer as a function of temperature.

Vol. 2 No. 12 Dec. 2004 31

Selected Papers

Mach-Zehnder interferometer. (QBER) was estimated from the difference between

2.4.2 Experimental setup the created keys. The polarization state was set to pro-
Figure 6 shows the experimental setup. The trans- duce the worst QBER in this measurement to test the
mitter, Alice, randomly phase-modulated a pulse worst-case scenario. The raw key generation rate was
train of coherent light by {0, π} for each pulse. The estimated from the number of photons counted and
pulse train was created by intensity-modulating con- the time over which the time interval analyzer (TIA)
tinuous wave (CW) light from an external-cavity actually measured. The results are shown in Fig. 7,
laser diode (wavelength: 1551 nm). The pulse width where the QBER and the raw key generation rate are
was 125 ps and the repetition rate was 1 GHz. The plotted as a function of the time window, which here
phase modulation was imposed with a LiNbO3 mod- is the margin for the photon arrival time. Since our
ulator driven by a pulse pattern generator. Then, the APD and TIA had some timing jitter, the measured
light power was attenuated to be 0.1 photon per pulse arrival time fluctuated and errors may have been
on average and was injected into a 20-km fiber (prop- induced. To reduce this timing error, Bob took data
agation loss: 4.46 dB). The receiver, Bob, measured within the time window, at the expense of the key
the phase difference between two sequential pulses generation rate. When the time window was 0.6 ns,
using a PLC Mach-Zehnder interferometer, which key creation was performed at a rate of 3076 bit/s
was set to provide the best extinction ratio for the with a QBER of 5.0%. A sufficient QBER was
worst polarization state. Photon detectors were obtained to create a secret key after error correction
placed at the two outputs of the interferometer. With and privacy amplification.
an appropriate phase in the interferometer, detector 1
clicked for 0 phase difference between two consecu- 3. Conclusion
tive pulses and detector 2 clicked for π phase differ-
ence. Avalanche photodiodes (APDs) were used as After reviewing quantum cryptography technolo-
the photon detectors, which were gated at 5 MHz. gies, we described our QKD protocol called differen-
The gate pulse was synchronized with the light pulse. tial-phase-shift QKD. This scheme makes use of the
The quantum efficiency was about 4.24% and the uncertainty in the photon detection time. It has a sim-
dark count probability was about 2.21 × 10–5 per gate. ple configuration for practical implementation and is
In detecting photons, Bob recorded the photon arrival suitable for fiber transmission provided a stable
time and which detector clicked. Using the above polarization-insensitive interferometer is available.
setup, Alice and Bob created raw secret keys at their We showed that a PLC interferometer based on silica
respective sites, following the protocol described in waveguide technologies provides polarization-
the previous section. The quantum bit error rate insensitive operation stable enough for the DPS-

Clock
Alice
Pulse
Pulse
pattern
generator
generator
1 GHz 0.1 photon
1 GHz per pulse
External Intensity Phase
Attenuator
cavity laser modulator modulator
(λ = 1551 nm) (LiNbO3) (LiNbO3)
0 π π
1 GHz
125 ps Bob
Glass-waveguide Pulse
Mach-Zehnder generator
interferometer
5 MHz
Optical fiber Time
Det interval
Det analyzer
20 km (APD) (TIA)
(4.46 dB) 20 cm
(2.64 dB)

Fig. 6. Experimental setup.

32 NTT Technical Review

 Selected Papers

8.5
4000

7.5
Raw key generation rate (bit/s)
3000
6.5

QBER (%)
2000 5.5

4.5
1000

3.5
0
0.2 0.4 0.6 0.8 1

Time window (ns)

Fig. 7. QBER and the raw key generation rate as a function of the time window.

QKD in practice. In a transmission experiment over a 022317, 2003.

[7] T. Honjo, K. Inoue, and H. Takahashi, “Differential-phase-shift quan-
20-km fiber, we obtained a key generation rate of tum key distribution experiment using a glass waveguide interferom-
3076 bit/s and a QBER of 5.0%. These experimental eter,” CLEO/IQEC 2004, ITuJ1, 2004.
results indicate that practical implementation of the [8] A. Himeno, K. Kato, and T. Miya, “Silica-based planar lightwave cir-
cuits,” IEEE J. Sel. Top. Quantum Electron, Vol. 4, p. 913, 1998.
DPS-QKD scheme should be possible. [9] Y. Inoue, H. Takahashi, S. Ando, T. Sawada, A. Himeno, and M.
Currently, the key generation rate is limited by the Kawachi, “Elimination of Polarization Sensitivity in Silica-Based
performance of the photon detector. A higher rate Wavelength Division Multiplexer Using a Polyimide Half Wave-
plate,” J. Lightwave Technol., Vol. 15, p. 1947, 1997.
should be possible by improving the photon detector
performance. We plan to develop a high-performance
photon detector and use it in an experiment to show
the DPS-QKD’s potential for a high key generation
rate compared with other conventional QKD
schemes. Toshimori Honjo
Research Engineer, Optical Science Laborato-
ry, NTT Basic Research Laboratories.
4. Acknowledgment He received the B.S. degree in information sci-
ence and M.S. degree in mathematical and com-
puting sciences from Tokyo Institute of Technol-
We would like to thank Dr. H. Takahashi at NTT ogy, Tokyo in 1996 and 1998, respectively. In
1998, he joined NTT Software Laboratories.
Photonics Laboratories for providing the PLC Mach- Since 2003, he has worked at NTT Basic
Research Laboratories. His current work is
Zehnder interferometer. research on quantum information technologies
and quantum cryptography.

References
[1] N. Gisin, G. Ribordy, W. Tittel, and H. Zbinden, “Quantum cryptog-
raphy,” Rev. Mod. Phys., Vol. 74, p. 145, 2002. Kyo Inoue
[2] C. H. Bennett and G. Brassard, “Quantum cryptography: public key Senior Research Engineer, Optical Science
distribution and coin tossing,” Proc. Internat. Conf. Computer Sys- Laboratory, NTT Basic Research Laboratories.
tems and Signal Processing, pp. 175-179, 1984. He received the B.S., M.S., and Ph.D. degrees
in applied physics from the University of Tokyo,
[3] C. H. Bennett, “Quantum cryptography using any two nonorthogonal Tokyo in 1982, 1984, and 1997, respectively. He
states,” Phys. Rev. Lett., Vol. 68, p. 21, 1992. joined NTT Laboratories in 1984 and conducted
[4] A. K. Ekert, “Quantum cryptography based on Bell’s theorem,” Phys. research on optical communications. From 2001
to 2003, he was a visiting scholar at Stanford
Rev. Lett., Vol. 67, p. 6, 1991. University, USA, where he worked on quantum
[5] A. Muller, T. Herzog, B. Huttner, W. Tittel, H. Zbinden, and N. Gisin, optics.
“‘Plug & play’ systems for quantum cryptography,” Appl. Phys. Lett.,
Vol. 70, pp. 793-795, 1997.
[6] K. Inoue, E. Waks, and Y. Yamamoto, “Differential-phase-shift quan-
tum key distribution using coherent light,” Phys. Rev., Vol. A 68, p.

Vol. 2 No. 12 Dec. 2004 33

View publication stats