IMRE LAYOUT_Layout 1 8/1/13 12:04 PM Page 28

QUANTUM COMMUNICATIONS

Quantum Communications:
Explained for Communication Engineers
Sándor Imre, Budapest University of Technology

ABSTRACT progress in communications requires not only

increased bandwidth in the physical layer, but
Moore’s Law dictates that during the forth- also more sophisticated and faster algorithms.
coming decades, new communications and com- Richard Feynman was among the first to note
puting technologies that exploit quantum that if binary information can be represented
mechanics will become generally available. To physically, for example, by a burst of photons or
assist Communications Society members with the aggregated charge of a collection of elec-
this transition, we provide a pragmatic descrip- trons, quantum mechanical properties such as
tion of quantum communications theory that the polarization of a photon or the spin of an
introduces the quantum mechanical properties electron might also be exploited to encode and
of polarization and entanglement, which do not transmit information.
have analogs in classical communications theory. If we are able to utilize quantum mechanics
We show plausible examples of how these quan- for communication and computing purposes, we
tum properties may be exploited to create noise- would have the opportunity to construct more
free channels and to improve channel utilization efficient algorithms than their best classical
for distributed MAC protocols. Finally, we counterparts, thanks to physical phenomena
describe how entanglement is used to enhance available only in the quantum world. The path is
security and improve key distribution efficiency open to exceed the classical capacity of commu-
in symmetric-key cryptography systems used by nication channels, and the potential exists to
operational quantum cryptographic networks. improve communication protocols; for example,
questions and risks regarding the security of
INTRODUCTION public key cryptography can be eliminated in
global-scale networks.
Moore’s Law was born in 1965 as a claim that Whether classical or quantum, computing and
the number of transistors per chip roughly dou- communications have many overlapping areas.
bles every 18–24 months. On one hand, this For example, signal processing is critical for effi-
results in faster and faster processors fulfilling cient high-quality communication between par-
the dreams and expectations of academia, indus- ties, but it can also be classified as a computing
try, and consumers. On the other hand, there is topic. Quantum computing can be exploited to
a critical trade-off: we must etch thinner and provide computationally more efficient solutions
thinner lines on the surface of semiconductor to classical signal processing problems. Quan-
chips, which results in smaller and smaller tran- tum-assisted communications combines quantum
sistors. Finally, we reach the nanoscale (<10–9m) signal processing with classical communication
world, which is governed by rules and laws col- channels. Quantum-based communications is
lectively called quantum mechanics. These differ identified with the case of the channel itself
significantly from those of macro-world classical obeying quantum rules. This article focuses on
physics with which we are familiar through our quantum-based communications.
perception and traditional courses of study. To provide an engineering-oriented introduc-
Some of these new principles are similar to tion to quantum communications, this article is
classical macroscopic principles; others have no organized as follows. In this section we provide
analog. In any case, we must accept them as a motivation for communications technologists to
new framework for computing and communica- learn quantum communication basics because
tions, and figure out how we may exploit them Moore’s Law predicts that a transition into prac-
to improve communications technologies. tice may occur in the next decade. We lead the
Moore’s Law, which is based on observations reader across the border between the classical
related to microprocessors, comes from comput- and quantum domains with a pragmatic explana-
er science. However, the convergence of tion of relevant quantum theory, supplemented
telecommunications and computer science result- with some simple quantum communications
ed in digital information processing at all levels examples. Next, we explain entanglement, a
of the protocol stacks. Telecommunication nodes strange but very promising phenomenon with
such as routers and mobile base stations contain potentially revolutionary applications to commu-
plenty of processors. Therefore, the continuous nications. We present a currently available quan-

28 0163-6804/13/$25.00 © 2013 IEEE IEEE Communications Magazine • August 2013

IMRE LAYOUT_Layout 1 8/1/13 12:04 PM Page 29

tum communication application, Secure Quan-

tum Communications, that connects quantum
mechanical theory with everyday practice. Final-
ly, we conclude and summarize future directions.

TURNING COMMUNICATIONS FROM

CLASSICAL TO QUANTUM
The behavior of elementary particles, and thus
any communication or computing device built
from them, is subject to special rules called the
postulates of quantum mechanics. In the quantum
world, the postulates are analogous to the axioms
of Euclidean geometry. Instead of using theoret-
ical physics to describe the postulates, we
describe them in terms of pragmatic thought-
experiment examples [1] that may be more
immediately useful for understanding the engi- Figure 1. Horizontal and vertical polarization of light.
neering aspects of quantum communications.
First, we introduce three actors: Alice, Bob,
and Eve. They will send, receive, or even eaves-
drop information according to the current sce- pV
nario. Furthermore, we briefly review the
differences between the classical and quantum
nature of light. Classically, light belongs to the -p-45°
family of electromagnetic waves having a special p45°
a
property characterized by polarization. Polariza-
tion is the orientation of oscillations in the plane
perpendicular to a transverse wave’s direction of
travel. Although quantum mechanically light
behaves as a bunch of elementary particles a
(called photons), it preserves the polarization pH
property, but polarization in this latter case
obeys the postulates of quantum mechanics, as
seen later in this section.
p-45°
FIRST STEP: THE SIMPLEST
CLASSICAL SOLUTION
Let us start with a very simple communication
example that crosses the border between classi- Figure 2. Vector representation of different photon
cal and quantum methods of communications. polarizations.
Alice wants to send several bits of information
over an optical fiber to Bob. Classically, she can
easily encode information: for logical 1 she emits decode the logical values of arriving photons by
light on a certain wavelength (i.e., sends a burst determining their polarization. Note that Alice
of photons in a given timeslot), while for logical and Bob are still communicating in a classical
0 she sends nothing. Nowadays we are able to way.
produce and detect one single photon; therefore,
Alice can use one photon to deliver one bit of THIRD STEP: FULLY QUANTUM OPERATION
information. Alice knows that light can be polarized, not only
at 0° and 90°, but at any angle. She prepares a
SECOND STEP: CROSSING THE photon with 45° polarization, depicted in Fig. 2
CLASSICAL/QUANTUM BORDER as p45°, then sends it to Bob. Bob’s photon detec-
tor1 — often called a measurement device — is
Alice exploited the location property in the pre- able to decide whether a vertically or horizontal-
vious step (i.e., whether or not a certain photon ly polarized photon has been sent by Alice (i.e.,
is positioned in a given time slot). Now she uses only orthogonal angles are allowed; it is not pos-
another property, polarization, to encode infor- sible to detect any intermediate angles). Because
mation. the vector p 45° is equidistant from p V and p H ,
We depict horizontal and vertical polarization that is, the projections of p45° onto the axes of pV 1 A measurement device

in Fig. 1, and the corresponding vector represen- and p H have the same magnitude a (Fig. 2), it receives a photon in its
tation in Fig. 2, where vertical polarization is can be expressed by means of a linear combina- input and outputs logical
denoted by vector PV and horizontal polarization tion of p V and p H : p 45° = ap V + ap H . If Alice 0 or 1 depending on the
by vector PH in the perpendicular plane, respec- sends several photons, all of them prepared in angle between the axis of
tively. Alice chooses to represent logical 1 with p45°, Bob’s receiver will answer logical 0 in one photon’s polarization and
vertically polarized photons and logical 0 with half of the cases and 1 in the other half, and it the two built-in orthogo-
horizontally polarized photons. By using a pho- will do this randomly; hence, in the classical nal measurement axes.
ton polarization device, Alice sends information sense, there is no information being sent by The measurement axes
to Bob. The receiver on Bob’s side can readily Alice to Bob. This randomness is a uniquely are labeled 0 and 1.

IEEE Communications Magazine • August 2013 29

IMRE LAYOUT_Layout 1 8/1/13 12:04 PM Page 30

tails representing logical 0 and 1, can be

regarded as a classical bit emulator. If heads
is facing upward, the bit value is set to 0;
0 0 1 0 1 similarly, tails sets the bit value to 1. Alice
Alice or Bob or
turns the coin upward according to the bit
1 Flip 1 0 1 0
value she wants to communicate. Then she
hands it over to Bob, who decides which bit
Bit-flip channel
value is communicated by observing heads
or tails using his receiver (i.e., his eyes).
Third step: Fully quantum operation: Now, let
us consider the case when Alice and Bob are
0 p45° p45° 0 connected via a simple but noisy channel,
Alice Bob depicted in the middle of Fig. 3a. The so-
1 p-45° p-45° 1
Flip called noisy bit-flip channel model consists
of a switch that connects the channel input
to either the upper path or the lower path of
Bit-flip channel the channel model. The upper path is a
“shortcut” or “wire,” that is, an identity
Figure 3. a) Classical; b) quantum communication between Alice and Bob transformation leaving the input state
over a bit-flip channel. unchanged; while the lower path negates
(flips) the input bit: 0  1. Both the upper
identity path and lower flip path are, by
quantum feature that emphasizes the indetermi- themselves, deterministic. If the switch was
nacy of quantum mechanics. Fine, but what can set permanently to the upper path (the wire),
Alice and Bob conclude from this communica- Bob would have to do nothing to receive the
tion experiment? The photon prepared in p 45° sent bits correctly. Conversely, if the switch
carries both possible logical values 0 and 1 at the was set permanently to the lower flip path,
same time. Now, this phenomenon is fully quan- to receive the sent bits correctly, Bob would
tum. have to perform the inverse of the flip trans-
Note: Although the representation of a formation, which is another flip transforma-
(quantum) one-qubit state in Fig. 2 and that of a tion. However, in the case of our bit-flip
(classical) M-ary modulation may appear to be channel model, the switch selects randomly
the same, there are significant differences. M-ary (with probability 1/2) between the upper and
modulations are visualized by means of complex lower paths (i.e., it introduces noise).
baseband equivalent, where each constellation Due to this randomness, Bob does not know,
point is represented by a single complex number: for a given bit, which path was selected by the
the axes of such figures denote the real and channel (switch); hence, Bob cannot build a clas-
imaginary parts (in phase and quadrature phase). sical receiver that is able to correct both effects
The distance of a constellation point from the — identity and flip — simultaneously. There-
pole (the magnitude of the vector) can be arbi- fore, Bob observes random noise, and cannot
trary since it is related to the signal strength. determine what Alice originally sent.
Thus, a classical M-ary modulation representa- Next, we describe how quantum operations
tion has two parameters (the real and imaginary can be exploited to make the above communica-
parts) with arbitrary magnitudes. In contrast, a tion noise-free. Returning to our coin-based
‘quantum constellation point’ in Fig. 2. is defined explanation, Eve stands between Alice and Bob,
by means of two complex numbers, i.e., we have and plays the role of the bit-flip channel. Similar
2 ¥ 2 = 4 parameters. (In Fig. 2. the two imagi- to the second step, Alice prepares her coin by
nary parts are assumed to be zero for the sake of facing it upward to show heads (0) or tails (1)
remaining visualizable.) Furthermore, we have and gives it to Eve. Randomly and with proba-
the condition that only unit-magnitude vectors bility 1/2, Eve either leaves the coin unchanged
are allowed. Finally, if we increase the number or turns it upside down (i.e., flips it), and then
2 Many are familiar with of bits/qubits the M-ary figure remains two- hands it to Bob. As before, because of Eve’s ran-
Einstein’s famous quote: dimensional, while each new qubit doubles the dom flip operation, Bob cannot be sure which
“God doesn’t play dice number of parameters. bit value was sent originally by Alice.
with the world,” but there To explain how this random-generator proto- In order to improve the communication,
are few who are familiar col may be useful for communications, we first Alice tosses the coin in the air in such a way that
with Max Born’s response introduce an everyday analogy — a coin 2 — of it rotates clockwise if she has a 0 to transmit and
to this: “If God has made this communication system in order to simplify counterclockwise if she has a 1 to transmit. Eve
the world a perfect the behavioral description of the protocol in the — the channel — catches the coin and randomly
mechanism, He has at quantum domain. The reader can therefore either leaves it unchanged or flips it. However,
least conceded so much to model and reproduce the experiments. Using the Eve will not change the direction of rotation, so
our imperfect intellect that coin analogy, the communication protocols she tosses the coin in the air with the same rota-
in order to predict little described above are now cast as follows: tional direction as she received from Alice.
parts of it, we need not First step: The simplest classical solution: If To detect the information encoded into the
solve innumerable differ- Alice wants to communicate 1, she gives a tossed coin, Bob constructs a receiver that is able
ential equations, but can coin to Bob in a given timeframe; other- to distinguish the two possible rotational direc-
use dice with fair suc- wise, to communicate 0, she gives nothing tions encoding 0 and 1 (e.g., a web camera that
cess.” A tossed coin is to Bob. can slow down the captured video taken from the
nothing else than the most Second step: Crossing the classical/quantum coin so that Bob’s eyes can distinguish the rota-
simple two-faced dice. border: A two-sided coin, with heads and tional direction). Since Eve’s random flip action

30 IEEE Communications Magazine • August 2013

IMRE LAYOUT_Layout 1 8/1/13 12:04 PM Page 31

does not change the direction of rotation, Bob In order to place the above novelties in a
will detect the information from Alice without practical engineering context, we mention that in Near-future
any error. Furthermore, Alice and Bob did not general, both redundancy-free and noise-free
experiments target
apply any redundancy; that is, the error correc- communication cannot be achieved. Therefore,
tion was implemented without extra coins (bits). classical error correction techniques have been Earth-satellite free
The aforementioned quantum mechanical extended to quantum channels. Various block space quantum
properties, which do not have analogs in classical coding methods were developed [7] so that
communications theory, are called out by Fig. 3. redundancy improves the transmission, while communications.
There is no classical error correction code that pilot-symbol-based solutions — where reference For all the media
can overcome a p = 0.5 bit-flip channel. Redun- bits are sent prior to payload data to measure
supporting quantum
dancy does not help. Using quantum codes, we the channel behavior — have recently appeared
can adjust the input states in such a way that in the literature [8]. These techniques are beyond communications,
they will be invariant to the channel effect. the scope of this article. however, the current
Now, we return to the original track at the As a conclusion to this section it is worth
third step where Alice and Bob use polarized highlighting the essence of entering the quantum lengths of the
photons to communicate. There, we concluded domain. The advantage comes from the connections are
that sending p45° photons will result in random increased degrees of freedom. In general, if we
limited by the lack of
results at Bob. replace the conventional classical physical media
Let us assume that Alice is connected to Bob representing logical values 0 and 1 with nano- effective quantum
via a quantum bit-flip channel that randomly scale objects (photons, electrons, etc.) that are repeaters.
either leaves each photon’s polarization governed by quantum phenomena, we are
unchanged or exchanges the coefficients of each allowed to use linear combinations of the logical
photon’s polarization components representing values. For example, for a polarized photon p =
the logical values (0  1); that is, if originally p apv + bpH, coefficients a and b can be adjusted
= ap v + bp H , applying an exchange operator to optimize any communication protocol; thus,
exch(◊) (called the Pauli-X operator in the litera- they represent new degrees of freedom that have
ture) gives exch(p) bpv + apH. no analogs in the classical domain.
According to our coin example, Alice encodes Before continuing, it is important to note that
classical 0 and 1 values into p45° and p–45° pho- quantum information can be delivered over vari-
tons, respectively. The tossed coin rotating in the ous media. One channel is optical fibers fed by
air emulates a quantum bit (below we will point polarized photons instead of light pulses. A
out what is the difference between a rotating practical obstacle, though, to using the current
coin and a p45° photon). It contains both heads fiber optic infrastructure is that commercial
and tails at the same time when rotating, similar fibers are engineered to operate at different
to a p45° photon containing pv and pH polariza- wavelengths than those that are optimal for
tions in an equally weighted linear combination quantum information transmission. Early trials
p45° = apv+ bpH. exceed 200 km distance. Free space quantum
Next we investigate the effect of the quantum channels are also available, bridging transmitters
bit-flip channel on the sent photons by means of and receivers located more than 100 km apart.
Fig. 2. The channel either lets the p in = p 45° Near-future experiments target Earth-satellite
photon through without any modifications, pin = free space quantum communications. For all the
p out = p 45°, or exchanges the coefficients of p v media supporting quantum communications,
and p H : ap v + ap H  ap H + ap v  ap H + ap v . however, the current lengths of the connections
Because the weights are equal, the photon polar- are limited by the lack of effective quantum
ization remains unchanged: pout = apH + apv = repeaters.
p45°. In the case of pin = p45° = –apv + apH pho-
tons, again the channel either performs an iden-
tity transform pout = pin = p–45° or modifies the ENTANGLEMENT:
photon to p out = –ap H + ap v = –(–ap v + ap H)
= –p–45° .
A STRANGE RESOURCE
Now, Bob applies a polarization detector In the previous section we have already crossed
adjusted to p ±45°. This device is able to distin- the border between the classical and quantum
guish photons according to their polarization domains, and learned that some quantum phe-
angles, and it ignores the sign of the polarization nomena can be exploited to improve communi- 3 If this article focused on
vector. Therefore, the detector will show 0 to cations. In this section, we describe and exploit quantum computing (vs.
Bob if a p45° photon has been received and 1 for another quantum phenomenon called entangle- quantum communica-
each ±p–45° photon. Since the channel does not ment. Again, we use a practical example to intro- tion), we would call the
modify the polarization angle of the sent pho- duce entanglement instead of citing the pure readers’ attention to the
tons, Bob decodes the information without any theory. fact that two coins
error: this classical-to-quantum encoding method Let us again use the coin-tossing analogy, but (“qubits”) can encode 22
makes the communication noise-free. now Alice uses two coins. She tosses both coins = 4 integers; therefore, n
Two aspects have to be emphasized when into the air, one after the other, and asks Bob to = 500 qubits comprises
evaluating this quantum protocol. First, we catch them (i.e., forcing them into heads or tails more classical numbers
improved the classically zero capacity of the bit- positions). Bob will receive one of the bit-pairs than the number of all the
flip channel to maximal capacity. Second, as 00, 01, 10, and 11,3 since each coin can be either atoms in the known uni-
mentioned above, we used no redundancy for 0 (heads) or 1 (tails) in Bob’s palm. verse. The readers may
error correction. Neither aspect would be possi- Next Alice glues one end of a matchstick to imagine the parallel infor-
ble in the case of classical communications. To the first coin and the other end to the second mation storage capacity
achieve these improvements, we had to enter the coin so that both coins show heads upward. This offered by quantum com-
quantum domain. results in synchronized rotation of the two coins puting.

IEEE Communications Magazine • August 2013 31

IMRE LAYOUT_Layout 1 8/1/13 12:04 PM Page 32

after tossing. Because of the matchstick, which entangle only elementary particles (photons,
Many cyber attacks implements a strict connection between the two electrons, etc.) and some larger structures, but
coins, if Alice catches one of the coins, the other always below the nano-scale. At present, we do
are undetected, or coin will also stop rotating. Therefore, Alice will not know how to entangle macroscopic objects
not detected until receive either 00 or 11. such as coins or space ships. Furthermore, the
well after serious To make the game more relevant to telecom- result of a coin tossing appears to be random,
munications, Alice uses a one-mile-long match- but we know that by using the appropriate for-
technical and stick 4 to connect the coins. Alice tosses them mulas of classical physics with carefully chosen
financial damage has and asks Bob to walk to the second rotating coin parameters, we are able to deterministically cal-
while the first is spinning in front of Alice. The culate the result. In contrast, according to our
occurred. Quantum first coin is collocated with Alice, while the sec- current knowledge about quantum mechanics,
optics-based ond is collocated with Bob. Now, Alice catches the measurement of a p 45° photon with a p v–H
solutions may her coin, and observes it in 0 (heads upward) detector will be fully random (i.e., no one can
position. What will Bob experience? Suddenly, predict the measurement outcome). This is why
provide new his coin stops rotating and remains in the 0 we mentioned above that coin tossing is only an
effective defense state. Similarly, if Alice observes 1, Bob will also emulation of what happens in the quantum
observe 1. For multiple coin tosses, 00 and 11 domain, but a very plausible emulator.
methods. will appear randomly. Entanglement plays an important role in
We can reproduce this experiment with two quantum mechanics, and therefore in quantum
photons by identifying heads (0) and tails (1) computing and communications. The noise in
with vertically and horizontally polarized pho- quantum communication channels can be traced
tons. A tossed coin rotating in the air represents back to the entanglement between the channel
both heads and tails at the same time, similar to and its environment. If a photon traversing an
a photon polarized at a 45° angle. Its polariza- optical fiber is entangled with another photon
tion is the linear combination of p v and p H as outside the fiber, and this latter photon has been
depicted in Fig. 2. When Bob measures this p45° subjected to some effects, these effects will influ-
photon with a pv–H measurement device, he will ence the state of the first photon in the fiber.
randomly get pv or pH (0 or 1). If Alice prepares Therefore, either we keep the photons from
two photons at a 45° angle one after the other becoming entangled, or we suffer the conse-
and sends them to Bob, Bob will detect 00 or 01 quences (channel noise). Fortunately, to entan-
or 10 or 11. gle two photons, they must be located very close
Next, Alice produces the analogy of the two to each other at some point in time; that is,
matchstick connected coins. She generates two entanglement cannot be created between distant
photons by means of a nonlinear beta-barium particles. Also, whereas naturally occurring
borate crystal so that their polarizations will be entanglements cause undesirable noise in quan-
connected [2], although this connection remains tum communications channels, man-made delib-
invisible. Alice keeps the first photon and sends erate entanglements may be exploited to improve
the second to Bob. Next, Alice measures her communications and computing capabilities.
photon with a detector. It behaves like a photon We end this section with a simple communi-
(i.e., Alice observes either 0 or 1 randomly). cation example where entanglement may be
Now, if Bob observes his photon anytime after exploited in future networks. Above, we consid-
Alice has observed her photon, Bob will observe ered a physical layer problem: how to mitigate
the same value as Alice did, because Alice’s or possibly eliminate noise in a communication
measurement also set Bob’s photon to the same channel. Now we step up to the medium access
value. Therefore, the result of the two measure- control (MAC) layer and describe how entangle-
ments will be either 00 or 11. ment may be used to improve the capacity uti-
This invisible binding effect is called entangle- lization of shared media. For exemplary
ment, referring to the fact that the two photons purposes, we use the ALOHA protocol, which is
have a special connection: acting on one of them a simplified version of the distributed MAC pro-
will influence the state of the other. Those read- tocol applied in WiFi networks.
ers who find this spooky action at a distance — as In an ALOHA-mediated system, the time
Einstein [3] called it — too surprising are asked dimension is divided into intervals (time slots)
to recall that by means of invisible electromag- with fixed duration (Fig. 4). At the beginning of
netic waves, remote devices can interact with each slot, Alice and Bob decide whether or not
each other. Unlike electromagnetism, however, to send a packet. If only one of them sends a
interactions between entangled photons occur packet into the channel, the packet will be deliv-
instantaneously, regardless of how far apart the ered. If neither of them sends a packet into the
photons are, whereas electromagnetic interac- channel, the slot will remain empty, and a trans-
tions are bounded by the speed of light. A discus- mission opportunity is lost. Finally, if both par-
sion of this amazing phenomenon is beyond the ties attempt to send their packets, a collision will
scope of this article; we mention it here not to occur, and again the transmission opportunity
confuse the reader, but to further distinguish the will be lost. In centralized MAC solutions, there
quantum domain from the classical domain and is a node (access point or base station) which
to make the reader aware of yet another quan- coordinates Alice and Bob so that collisions do
tum phenomenon that may be exploited some- not occur. However, there are many protocols
how in the future to improve communications. that use distributed MAC (e.g., ad hoc network-
Before introducing a communications appli- ing protocols and sensor network protocols).
4This is only emulation; cation example that exploits entanglement, we According to information theory, the best dis-
nature will be more sur- pause briefly to explain one difference between tributed solutions (as measured by fairness and
prising, as we will discuss. a photon and a coin. Currently we are able to overall throughput) have these properties:

32 IEEE Communications Magazine • August 2013

IMRE LAYOUT_Layout 1 8/1/13 12:04 PM Page 33

• Alice and Bob have random access to the

channel. OK OK Unused Collision
• The probability of each user sending a pack- Alice
et in a certain slot is 1/N, where N is the
number of active users.
Time
This means that, taking into account the possible Bob
four outcomes for the case of Alice and Bob
(two users), only half of the slots will be utilized.
Figure 4. Slotted ALOHA medium access.
This is inefficient.
Centralization would improve utilization, but
this approach may not be possible, practical, or
desirable in many situations. However, let us SECURE QUANTUM
consider that the system attribute necessary to
improve utilization is coordination between COMMUNICATIONS: A READY-FOR-
users. Centralization enables coordination — but MARKET APPLICATION
are there other approaches besides centraliza-
tion that will enable coordination? Is it possible Security — authentication, cryptography, priva-
to utilize quantum communications to achieve cy, and so on — is now a requirement for com-
coordination among distributed users? munication networks. Today, however, attack
Let us consider an entangled photon pair that defenses are insufficient and ineffective against
behaves like the coins stuck together with a the current cyber threat. Many cyber attacks are
matchstick. The only modification is that one of undetected, or not detected until well after seri-
the coins (photons) is negated, that is, an upward ous technical and financial damage has occurred.
heads coin is connected to an upward tails one. Quantum optics-based solutions may provide
Logically this encodes 01 or 10. First, Alice and new effective defense methods.
Bob prepare such a connected pair and share We focus here on cryptography, which is
them. Later, when they want to communicate, the responsible for secure communications in the
photons have to be measured. To avoid collisions, presence of third parties [4]. In our example
the measurement devices at Alice and Bob select Alice wants to send information — called plain
randomly between the possible 01 or 10 results; text or plain message — securely to Bob. Eve is
whoever gets the 1 is allowed to send the packet, interested in eavesdropping this information. In
and the other with 0 must remain silent. This pro- the communications context, two types of cryp-
tocol remains random, fair, and distributed, but tography are of interest: symmetric-key and pub-
provides perfect coordination between the parties. lic-key.
At first glance, it may appear that there is In the case of symmetric-key cryptography,
nothing more being done here than could be Alice ciphers the plain text by means of a special
done by a shared pseudo-random number gener- binary code, referred to as the key. Bob has a
ator (if the next bit is 0, Alice transmits; other- copy of the key, and he uses it to decipher the
wise, if it is 1, Bob transmits). However, it would received message to the plain text. Because they
not work in a distributed environment since a use the same key for ciphering and deciphering,
central entity is required that generates the ran- the method is called symmetric. We like sym-
dom codes and distributes them among the metric-key cryptography because Shannon
users. In contrast, the quantum approach does showed that an ideal symmetric-key cryptograph-
not need such a central entity since entangle- ic system, called a one-time pad or Vernam
ment can be generated in a distributed way, that cipher, is provably secure, and furthermore, it is
is, there is no need for central entity selection the only cryptographic system that is secure. In a
prior to code generation. Another difference is Vernam cipher, the key is:
that quantum mechanics provides truly random (a) The same length as the plain text
numbers instead of pseudo-random ones. (b) Truly random
This means that entanglement enables coor- (c) Never reused
dination among distributed users — and there- (d) Kept secret
fore improves utilization — without resorting to Although all of these key properties are achiev-
centralization. able in practice (with varying degrees of difficul-
Entanglement is often used in emerging quan- ty) using classical methods, the secrecy property
tum communications (and computing) systems. (d) is particularly difficult to achieve in today’s
One of its most important applications is increas- Internet environment. Below, we show how
ing the capacity of a channel by entangling con- quantum methods can significantly mitigate the
secutively sent bits [11]. The significance of difficulty.
communications protocols that exploit entangle- The primary issue with secrecy is that Alice
ment is evidenced by the definition of a new and Bob have to meet to share the key before
capacity measure: entanglement-assisted capacity. starting the communication. There is an effective
As a most extreme application of entanglement, but expensive solution to this problem: central-
we mention that there are quantum channels ized infrastructure. For example, in the case of
which individually have zero information capaci- cellular mobile networks, for each user (Alice or
ty, but by exploiting entanglement they can be Bob), two copies of the key are stored:
superactivated: by combining two such channels, 1. In the user’s SIM card
they can deliver non-zero information to the 2. In the network database
receiver [12]! It is not surprising that entangle- To communicate with Bob, Alice first establishes
ment is often called a resource in the quantum symmetric-key communication with the network.
communications and computing literature. A dedicated network node, which functions as

IEEE Communications Magazine • August 2013 33

IMRE LAYOUT_Layout 1 8/1/13 12:04 PM Page 34

Although quantum computing is out of scope

Trial Status for this article, we mention that certain quantum
algorithms efficiently solve this factorization
10-node quantum key distribution network in problem [5, 6]. Therefore, the first desktop
DARPA Quantum network quantum computers will significantly influence
the United States since 2004.
the evolution of asymmetric key systems that
200 km of standard fiber optic network inter-
depend on the computational difficulty of factor-
SECOQC ization. Symmetric-key systems may therefore
connecting six locations in Austria, 2008.
displace asymmetric systems in the market if the
centralized infrastructure bottleneck can be
Two-year reliability test of commercial encryp-
eliminated.
SwissQuantum tion technology in Geneva, Switzerland,
metropolitan area, 2009–2011.
Fortunately, quantum communication-based
cryptography can handle not only classical
eavesdroppers but also hackers using quantum
45 km four-node testbed in Japan launched in
Tokyo QKD Network computers. Furthermore, while we may have
2010.
to wait decades for practical quantum comput-
Table 1. Major commercial QKD trials. ers, the first prototypes of quantum crypto-
graphic networks [10] and products have been
under development for nearly a decade and
are ready for market; see Table 1 as well as
an intermediary back-to-back user agent, first the products available from MagiQTech,
deciphers the message with the copy of Alice’s idQuantique, and QuintessenceLabs. There
key, enciphers the message with Bob’s key, and are several classes of quantum cryptographic
then sends the (enciphered) message to Bob. protocols, but the basic concept is similar. If
Bob deciphers the message to plain text. This Alice were able to send a copy of her key to
two-hop method achieves secrecy if the dedicat- Bob with the guarantee that any eavesdrop-
ed network node is kept safe. ping attempt will be revealed, they can use
However, this solution does not work in prac- symmetric-key cryptography, which has theo-
tice for distributed multiprovider networks such retically proven security: quantum cryptogra-
as the Internet. Historically, Internet users have phy removes the security “holes” in
never used centralized infrastructure, and are symmetric-key cryptography.
simply not going to use services that require cen- Entanglement can be exploited to securely
tralized registration. exchange the ciphering key [1]. Recall from
Asymmetric-key (or public-key) cryptogra- the beginning of the previous section that if
phy can help in these scenarios. Bob generates Alice produces an entangled photon pair which
a random key and keeps it private. Next, he contains 00 or 11 and sends one of the pho-
generates a public key, derived from the pri- tons to Bob, measuring the photons at both
vate key, by means of a special function f(). sides, they will both observe 00 or 11 random-
Then he announces the public key on the Inter- ly. From the cryptography point of view, they
net such that it can be accessed by anyone who generated the first bit of a symmetric cipher-
wants to communicate with Bob. Alice wants to ing key-pair since the value of the bit is fully
communicate with Bob, so she uses Bob’s pub- random and agrees at both parties. As
lic key to encipher the plaintext message, and explained in the previous section, no one can
then sends the ciphertext to Bob. Bob deci- influence the random behavior of photon mea-
phers with his private key and reads the plain- surement, so no one is able to predict the
text message. results of the measurement (i.e., Eve cannot
The practicality and security of public-key figure out the bits of the key).
cryptography depends on the function f() having If they repeat these steps several times, they
two properties. First, it has to be efficiently com- can produce a ciphering key of appropriate
putable so that Bob can produce the public key length. The reader may ask the following obvi-
easily. Second, the calculation of the inverse of ous question emerging from our classical percep-
f() should be computationally hard to prevent tion of nature: Who generated the key? The
Eve from being able to compute the private key answer might be surprising. Since Alice and Bob
from the public key. For example, for the widely did the same measurements at the same time,
used RSA algorithm, Bob efficiently multiplies the key was produced by both Alice and Bob. It
two large (secret) prime numbers to generate follows that the key did not exist before the
the public key. The private key is also generated measurements; thus, Eve had nothing to eaves-
from the two large prime numbers. Eve can drop. Therefore, quantum key distribution can
decrypt any messages that have been encrypted rather be regarded as distributed quantum key
using Bob’s public key, provided she can gener- generation.
ate the private key by factoring the public key In practical implementations QKD protocols
into the two prime numbers used to generate it. are vulnerable to man-in-the-middle attacks
However, there is no known efficient classical when a malicious eavesdropper presents to Alice
integer factorization algorithm, and because of as if he were Bob, and to Bob as if he were
the size of the prime numbers — typically 1024 Alice. To prevent such attacks, authentication of
bits, or over 300 decimal digits — the computa- the parties is needed, which requires pre-shared
tional resources necessary to factor the public keys. Once authentication has been performed
key are prohibitive. The security of this asym- successfully, it is possible along with quantum
metric-key cryptography method depends on the key distribution to expand this key to authenti-
second property of f(). cate the next session.

34 IEEE Communications Magazine • August 2013

IMRE LAYOUT_Layout 1 8/1/13 12:04 PM Page 35

CONCLUSIONS [4] A. J. Menezes, P. C. van Oorschot, and S. A. Vanstone,

Handbook of Applied Cryptography, Taylor & Francis
Inc., 1996, p. 816. State-of-the-art
Quantum mechanics is beginning to play an [5] P. W. Shor., “Polynomial-Time Algorithms for Prime Fac-
important role in many areas of everyday life, torization and Discrete Logarithms on a Quantum Com-
research activities
ranging from high-endurance materials to puter,” SIAM J. Computing, vol. 26, no. 5, Oct. 1997, focus, on one hand,
pp. 1484–1509.
pharmaceuticals, and now to communications, [6] L. Grover, “A Fast Quantum Mechanical Algorithm for on the theoretical
where it has multiple applications. We have Database Search,” Proc. 28th Annual ACM Symp. Theo-
reviewed how classical communications can be ry of Computing, May 1996, pp. 212–19. limitations of
extended to the quantum domain and how the [7] M. A. Nielsen and I. L. Chuang, Quantum Computation
and Quantum Information, Cambridge Univ. Press,
quantum channels
offered new phenomena can be exploited to 2000, p. 676.
improve communications systems. A special and protocols, and
[8] L. Gyongyosi, S. Imre, “Pilot Quantum Error-Correction for
resource called entanglement was explained. Noisy Quantum Channels,” Proc. 2nd Int’l. Conf. Quantum on the other hand,
In order to plant these novelties into engi- Error Correction, Los Angeles, CA, 5–9 Dec. 2011.
[9] S. Imre and L. Gyongyosi, Advanced Quantum Communica- practical implemen-
neering practice, we have discussed distribut- tions: An Engineering Approach, Wiley, 2012, p. 488.
ed medium access control and an application [10] C. Elliott, D. Pearson, and G. Troxel, “Current Status of
tations in the existing
providing secure key exchange between dis- the DARPA Quantum Network,” Comp. Commun. Rev., classical networks
tant parties. vol. 33, no. 4, pp. 227–38.
Many physical media can carry quantum [11] C. H. Bennett et al., “Entanglement-Assisted Capacity of a paving the way for
Quantum Channel and the Reverse Shannon Theorem,”
information, opening new horizons in communi- IEEE Trans. Info. Theory, vol. 48, no. 10, pp. 2637–55. spreading quantum
cations. State-of-the-art research activities focus, [12] T. Cubitt et al., “Superactivation of the Asymptotic Zero-
communications.
on one hand, on the theoretical limitations of Error Classical Capacity of a Quantum Channel,” IEEE Trans.
Info. Theory, vol. 57, no. 12, 2011, pp. 8114–26.
quantum channels and protocols, and on the
other hand, practical implementations in the
existing classical networks paving the way for BIOGRAPHY
spreading quantum communications [9]. SANDOR IMRE ([email protected]) received his M.Sc. degree
in electrical engineering from the Budapest University of
ACKNOWLEDGMENT Technology (BME) in 1993. Next he started his Ph.D. stud-
ies at BME and obtained a Dr.Univ. degree in probability
The author would like to thank Dr. Sean Moore theory and mathematical statistics in 1996, a Ph.D. degree
for his invaluable effort and suggestions, which in telecommunications in 1999, and a D.Sc. degree from
have improved the quality of this article. the Hungarian Academy of Sciences in 2007. Currently, he
is carrying on his activities as a professor and head of the
Department of Networked Systems and Services. He is
REFERENCES Chairman of the Telecommunication Scientific Committee
of the Hungarian Academy of Sciences. He was invited to
[1] S. Imre and F. Balazs, Quantum Computing and Communi- join the Mobile Innovation Centre as R&D director in 2005.
cations — An Engineering Approach, Wiley, 2005, p. 314. His research interest includes mobile and wireless systems,
[2] A. Zeilinger et al., Dance of the Photons, New York, and quantum computing and communications. He has
2010, p. 205. made contributions on different wireless access technolo-
[3] A. Einstein, B. Podolsky, and N. Rosen, “Can Quantum- gies, mobility protocols, security and privacy, reconfig-
mechanical Description of Physical Reality be Consid- urable systems, and quantum-computing-based algorithms
ered Complete?” Phys. Rev., vol. 47, 1935, pp. 777–80. and protocols.

IEEE Communications Magazine • August 2013 35