IEEE Communications Magazine November 2014152 0163-6804/14/$25.00 2014 IEEE

G. Zheng is with Universi-ty of Essex. He is alsoaffiliated with Interdisci-plinary Centre for Securi-ty, Reliability and Trust(SnT), University of Lux-embourg, Luxembourg

I. Krikidis and S. Timoth-eou are with University ofCyprus.

C. Masouros is with Uni-versity College London.

D.-A. Toumpakaris iswith University of Patras.

Z. Ding is with LancasterUniversity.

INTRODUCTIONResources (e.g. time, frequency, code) have tobe shared by multiple users in wireless networks.Therefore, interference has long been consid-ered as a deleterious factor that limits the sys-tem capacity. In conventional communicationssystems, the design objective is to allow users toshare a medium with minimum or no interfer-ence. Thus, great efforts are made to avoid, miti-gate, and cancel interference. For instance, tosupport multiple users, orthogonal access meth-ods in time, frequency, code as well as spatialdomains have been used in different generationsof cellular systems. In order to improve the cov-erage and the capacity in future-generation het-erogeneous networks that will likely contain alarge number of uncoordinated low-power nodessuch as femtocells, interference needs to be miti-gated in multiple domains, rendering its manage-ment a challenge.

Interference mitigation/avoidance techniquesprovide convenient mechanisms to allow multi-ple users to share the wireless medium. Howev-

er, they lead to inefficient use of wirelessresources. One may ask whether to cancel ormitigate interference is always the optimal wayof utilizing wireless resources. Indeed, there hasbeen growing interest in exploiting interferenceto improve the achievable rate, the reliability,and the security of wireless systems. Recently,new views on interference have resulted inadvanced interference-aware techniques, which,instead of mitigating interference, explore thepotential of using interference. We present twoexamples from the literature to illustrate the ideas.

In his early work on dirty paper coding [1],Costa proved the striking result that interferenceknown at the transmitter but not at the receiverdoes not affect the capacity of the Gaussianchannel. The optimal strategy to achieve thisinterference-free capacity is to code along inter-ference, while canceling interference is strictlysub-optimal. Another example is coordinatedmultipoint or multi-cell coordination, where basestations (BSs) cooperate to serve their own andout-of-cell users.

In the downlink, the cooperating BSs worktogether to jointly optimize the transmitterstrategies such as power, time, and beamformingdesign to control the inter-cell interference. Cell-edge users who suffer most from the inter-cellinterference now benefit most from this coordi-nation. In the uplink, joint decoding is per-formed in BSs, so signals from users in othercells are no longer treated as interference, but asuseful signals.

The purpose of this article is to re-examinethe notion of interference in communicationsnetworks and introduce a new paradigm thatconsiders interference as a useful resource. Wefirst give an overview from the information theo-retic standpoint as a justification for rethinkingthe role of interference in wireless networks. Wethen introduce interference alignment and signalalignment as effective means to handle interfer-ence and increase the achievable rates. Depart-ing from this traditional view, we present threenovel techniques of interference exploitationthat aim to improve the performance of wirelessnetworks. The first technique is a data-aidedprecoding scheme in the multiuser downlink thatjudiciously makes use of the interference amongusers as a source of useful signal energy. In the

ABSTRACTThis article re-examines the fundamental

notion of interference in wireless networks bycontrasting traditional approaches to new con-cepts that handle interference in a creative way.Specifically, we discuss the fundamental limits ofthe interference channel and present the inter-ference alignment technique and its extension ofsignal alignment techniques. Contrary to this tra-ditional view, which treats interference as adetrimental phenomenon, we introduce threeconcepts that handle interference as a usefulresource. The first concept exploits interferenceat the modulation level and leads to simple mul-tiuser downlink precoding that provides signifi-cant energy savings. The second concept usesradio frequency radiation for energy harvestingand handles interference as a source of greenenergy. The last concept refers to a secrecy envi-ronment and uses interference as an efficientmeans to jam potential eavesdroppers. Thesethree techniques bring a new vision about inter-ference in wireless networks and motivate aplethora of potential new applications and services.

ACCEPTED FROM OPEN CALL

Gan Zheng, Ioannis Krikidis, Christos Masouros, Stelios Timotheou, Dimitris-Alexandros Toumpakaris,

and Zhiguo Ding

Rethinking the Role of Interference inWireless Networks

ZHENG_LAYOUT.qxp_Layout 10/29/14 3:47 PM Page 152

IEEE Communications Magazine November 2014 153

second technique, we consider simultaneousinformation and energy transfer; in such a sys-tem, while interference links are harmful forinformation decoding, they are useful for energyharvesting. Thus, a favorable trade-off is demon-strated. The third technique leverages interfer-ence in physical layer secrecy as an effective wayto degrade the channel of the eavesdropper andincrease the systems secrecy rate.

INTERFERENCE FROM THEINFORMATION THEORETIC

STANDPOINT

We first present an overview of results on inter-ference from information theory. The Interfer-ence Channel (IC) models simultaneoustransmission by non-cooperating transmittersand receivers. The messages of each link areencoded only by the corresponding transmitter,and the receiver does not have access to the sig-nals of other receivers. Figure 1a depicts the K-userGaussian Interference Channel (G-IC). Each ofK transmitters wants to send a message to thecorresponding receiver. Receiver k bases itsdecision on signal Yk, which contains not onlythe (scaled) useful signal Xk, but also interfer-ence and Gaussian noise.

Despite its apparent simplicity, to this date itis not known what the optimal way of transmit-ting over the G-IC is, not even for the two-userG-IC shown in Fig. 1b. Nevertheless, significantprogress has been made recently, and resultsfrom information theory have started influencingthe design of wireless networks. The optimaldecoding strategy depends on the power of inter-ference compared to the direct links. Interfer-ence should be treated as noise when it is veryweak. The exact conditions for the two-user G-ICto be in the very weak regime can be found in [2].In information theoretic terms, the messages ofboth transmitters are private, since they are onlydecoded at the intended receiver. On the otherhand, when the power of the interfering signalexceeds the power of the signal of interest (stronginterference), the optimal strategy is to alsodecode interference at the receivers. In this case,both messages are public. If the power of theinterference exceeds an even higher threshold,the G-IC is in the very strong interference regimeand the rate that can be achieved by each link isthe same as if the interferer did not exist, that is,interference does not impact the achievablerates. Nevertheless, the receiver does need todecode interference in addition to the signal ofinterest. Clearly, there are costs associated withinterference-aware decoding. The receivers aremore complex, synchronization is essential, andeach receiver needs to estimate not only its ownchannel, but also the cross-channel coefficients.

The most challenging situation arises whenthe power of the interference is of the order ofthe power of the signal of interest. To this dateit is not known what the best way to transmitand decode is. A strategy that combines publicand private messages (the so called Han-Kobayashi scheme [2]) achieves higher ratescompared to treating interference as noise or

avoiding it via orthogonal transmission, orattempting to decode all messages at each receiv-er. Moreover, it has been shown that as the sig-nal-to-noise ratio (SNR) grows to infinity, asimplified Han-Kobayashi scheme can attain thecapacity region within 1/2 bit [2]. In addition toproviding evidence that strategies based on theHan-Kobayashi scheme may be the best for thetwo-user G-IC, this result may prove useful infuture wireless networks with small cell size thatwill operate at high SNRs and will therefore belimited by interference rather than by noise.

Devising good strategies for the K-user G-ICseems to be even more challenging, and theHan-Kobayashi scheme does not appear toextend to the K-user G-IC in a straightforwardmanner. A promising direction toward findinggood strategies for the K-user G-IC appears tobe dealing with the combined interference by allK 1 users at each receiver instead of decodingseparately the interference by each user. Fur-thermore, a deterministic approximation frame-work has been developed for the G-IC, whichenables the construction of structured codes [2].By employing structured lattice codes, which arealso used in other scenarios, such as multi-wayrelay channels, it is possible to attain the capaci-ty region of the G-IC within a constant gap [3].Very recently, there has been an interesting find-ing that connects topological interference man-agement and index coding [5]. This connectioncan be leveraged to calculate rate regions thatare within a constant gap from capacity and todevelop transmission schemes over wireless net-works. The existing index coding solutions arethen translated to interference managementsolutions via a family of elegant achievabilityschemes of interference alignment (IA) that hasgenerated significant interest, and is discussed inmore detail below.

System designs that operate based on the bestknown achievability schemes of information the-ory being the ultimate goal, in the meantimeimprovements in performance can also beattained by incorporating interference-awareschemes in current systems. In [6] it is shown

Figure 1. a) The K-user Gaussian Interference Channel (G-IC); b) the two-user G-IC.

(b)(a)

Y1

Z1

g11

g11

g12

g2K

g12

g22

g21

+

Y2

X1

X2

Z2

+

YKXK

ZK

+

Z2

Z1

Y2X2 +

... ...

Y1X1 +

ZHENG_LAYOUT.qxp_Layout 10/29/14 3:47 PM Page 153

IEEE Communications Magazine November 2014154

that when the transmitters use discrete constella-tions and interference-aware detectors areemployed at the receivers, the achievable ratesover the fading G-IC are limited by the SNRrather than by the signal-to-interference-plus-noise ratio (SINR).

INTERFERENCE ANDSIGNAL ALIGNMENT

Prior to the invention of IA [4], interferenceavoidance has been achieved by relying on theuse of orthogonal frequency or time channels.And when interference is inevitable, convention-al approaches are to adopt advanced decoding/detecting algorithms by treating interference asnoise.

The success of IA lies in the fact that it effi-ciently exploits the rich degrees of freedomavailable from the time/frequency/spatialdomains. By a careful coordination among thetransmitters, the use of IA can ensure that allthe interference is aligned together to occupyone half of the signal space at each receiver,leaving the other half available to the desiredsignal. As a result, the per-user rate achieved byIA for the interference channel with K pairs ofsingle-antenna transceivers is C(SNR) = 1/2log(SNR) + o(log(SNR)). This result is surprisingsince a traditional view is that such a K-user sce-nario is interference-limited and hence the per-user rate is diminishing by increasing the numberof users. As a result, the use of IA ensures thatthe spectral efficiency of wireless communica-tions can be improved significantly since moreusers sharing the same bandwidth yields a largersystem throughput.

In addition to the interference channel, theconcept of IA has also been applied to othercommunication scenarios, including the multipleaccess channel, the broadcast channel, theone/two-way relaying channel, as well as physicallayer security. In practice, the implementation ofIA is not trivial since global channel state infor-mation at each transmitter (CSIT) is required,which is challenging, particularly for the casewith fast time-varying channels. Two types of

approaches to realize IA in practice have beenproposed. One is to apply advanced feedbacktechniques, and existing results have demonstrat-ed that the number of fedback bits needs to beproportional to the SNR in order to achievenearly optimal performance [9]. The other is toexploit the coherent structure of channels andapply manipulations analog to space time codingat the transmitters. As a result, the concept ofIA can be implemented even when the channelinformation is not available to the transmitters.

The concept of signal alignment can beviewed as an extension of IA in the context ofbi-directional communications [10] and [11]. Forexample, consider a multi-pair two-way relayingcommunication scenario as shown in Fig. 2,where M pairs of source nodes exchange infor-mation with their partners via the relay. Eachsource node is equipped with N antennas, andthe relay has M antennas. As can be seen fromFig. 2, the relay observes 2M incoming signalstreams, and needs to have at least 2M antennasin order to separate these signals. The use of sig-nal alignment is to effectively suppress intra-pairinterference and reduce the requirement on thenumber of antennas at the relay. Particularly, bycarefully designing the precoding vectors at thesources, the intra-pair interference is aligned atthe relay, which means that the original 2M sig-nal streams are merged into M streams. As aresult, a relay with only M antennas can accom-modate 2M incoming signals, which is particular-ly important for practical scenarios where nodesare equipped with a limited number of antennas.At the user end, each receiver can first subtractits own information, the so-called self-interfer-ence, and then detect the information from itspartner, a method analogous to network coding.

DATA-AWARE INTERFERENCEEXPLOITATION FOR

MULTIUSER TRANSMISSIONThe a priori knowledge of interference is readilyavailable at downlink transmission, where CSITcombined with the knowledge of all data sym-bols intended for transmission can be used to

Figure 2. Illustration of the concept of signal alignment: a) system diagram for multi-way relaying, N M; b) precoding design toensure interference alignment.

Node 1

1

Node M

Node m

Node 1

u1s1u1s1

umsm

uMsM

umsm

H1u1+H1u1

Hmum+Hmum

HMuM+HMuM

uMsM

(b)(a)

Relay

H1H1

HMHM

HmHm

Node 1

Node m

Node M

2...N

M

2

1

Node MNode M

12...N

Node 1

12...N

.

.

.

.

.... .Phase 2

Phase 1

12...N

Phase 2

Relay

Phase 1

ZHENG_LAYOUT.qxp_Layout 10/29/14 3:47 PM Page 154

IEEE Communications Magazine November 2014 155

explicitly predict the resulting interferencebetween the symbols. Despite the insights in [1],the majority of existing precoding implementa-tions attempt to eliminate, cancel, or pre-sub-tract interference. Only recently, however, hasthere been growing interest in making use of theinterference power to enhance the useful signal[7, 8]. Indeed, it has been shown that interfer-ence can contribute constructively to the detec-tion of the useful signal, and this phenomenoncan be utilized in the CSIT-assisted downlinktransmission and other known-interference sce-narios to improve performance without raisingthe transmit power.

To clarify the above fundamental concept, atrivial example of two users is shown in Fig. 3a,where we define the desired symbol as x1 andthe interfering symbol as x2. Without loss of gen-erality we assume that these belong to a BinaryPhase Shift Keying (BPSK) constellation andthat x1 = 1, x2 = 1. For illustration purposes,we assume a lossless channel from the intendedtransmitter to the receiver and an interferingchannel represented by the coefficient r. Ignor-ing noise, the received signal is

y1 = x1 + x2 r, (1)

where x2 r is the interference. Note that thismodel also corresponds to multi-antenna trans-mission with matched filtering where the corre-lation between the two channels is r. In Fig. 3btwo distinct cases are shown, depicting the trans-mitted () and received (o) symbols for user 1 onthe BPSK constellation. In case i) with r = 0.5 itcan be seen from (1) that y1 = 0.5. The destruc-tive interference from user 2 has caused thereceived symbol of user 1 to move toward thedecision threshold (imaginary axis). The receivedpower of user 1 has been reduced and its detec-tion is prone to low-power noise. In case ii),however, for r = 0.5 (1) yields y1 = 1.5, andhence the interference is constructive. Thepower received has been augmented due to theinterference from user 2 and now its detection istolerant to higher noise power (nconstr comparedto north). It should be stressed that in both casesthe transmit power for each user is equal to one.Note that while the above example refers to atwo-user transmission scenario for illustrationpurposes, the fundamental concept can beextended to more users, multipath transmission,inter-cell interference, and other generic inter-ference-limited systems.

Clearly, there are critical gains to be drawnfrom the exploitation of constructive interfer-ence in interference-limited transmission. As afirst step, analytical constellation-dependentcharacterization criteria for systematically classi-fying interference to constructive and destructivehave been derived in [7, 8] and references there-in for PSK modulation. Early work carried outon a simple linear precoding technique hasreported multi-fold increases in the receivedSNR for fixed transmit power compared to zero-forcing (ZF) beamforming [7]. This can be non-trivially translated to multi-fold savings intransmit power for a fixed received SNR. A rep-resentative result is shown in Fig. 4a where therequired SNR per transmit antenna in a cellular

downlink for an uncoded symbol error rate(SER) of 102 is shown for increasing numbersof single-antenna users. The results compare thewidely known ZF precoding with the interfer-ence exploitation precoding of [7] for QPSK and8PSK modulation. Significant SNR gains of up to10dB (a 10-fold transmit power reduction) canbe observed between the two techniques, by simplyexploiting the existing constructive interference.

Further work has investigated the applicationof this concept on advanced nonlinear precod-ing, yielding further significant gains in thetransmit power. More recent work has extendedthis concept to inter-cell interference exploita-tion in multi-cellular transmission scenarios [8].The important feature in all the above tech-niques is that the performance benefits aredrawn not by increasing the transmit power ofthe useful signal, but rather by reusing interfer-ence power that already exists in the communi-cations system, a source of green signal powerthat with conventional interference cancellationtechniques is left unexploited.

WIRELESS INFORMATION ANDENERGY TRANSFER

Energy harvesting (EH) communications systemsthat can scavenge energy from a variety of natu-ral sources (solar, wind, etc.) for sustainable net-work operation have attracted significantinterest. The main limitation of conventional EHsources is that they are weather-dependent andthus not always available.

A promising harvesting technology that couldovercome this bottleneck is radio frequency(RF) energy transfer where the ambient RFradiation is captured by the receiver antennasand converted into a direct current voltagethrough appropriate circuits (rectennas). Theconcept of RF-EH is not new; over 100 yearsago, Nicola Tesla proved and experimentallydemonstrated the capability of transferring ener-gy wirelessly. The integration of RF-EH technol-ogy into communications networks opens newchallenges in the analysis and design of transmis-sion schemes and protocols. Multi-user interfer-ence, which is the main degradation factor inconventional networks, can be viewed as usefulenergy signals that could be exploited for har-vesting purposes. Although from an informationtheoretic standpoint the same signal can be usedfor both decoding and EH, due to practical

Figure 3. The concept of constructive interference a two-user example.

(b)(a)

i) =0.5

ii) =-0.5x1=1, x2=-1

y1=y1+x2

1

1

-1x1

x1

x2x1

y1

north

y1

nconstr

ZHENG_LAYOUT.qxp_Layout 10/29/14 3:47 PM Page 155

IEEE Communications Magazine November 2014156

hardware constraints, simultaneous energy andinformation transmission is not possible withexisting rectenna technology. Two practicalreceiver approaches for simultaneous wirelesspower and information transfer are time switch-ing (TS), where the receiver switches betweendecoding information and harvesting energy; andpower splitting (PS), where the receiver splitsthe received signal in two parts for decodinginformation and harvesting energy, respectively [12].

An interesting implication of the PS tech-nique is that in multiuser networks, harvestedenergy at a particular receiver can emanateeither from sources that intentionally transmittoward that direction or from other sourceswhose signal is traditionally perceived by thatreceiver as interference. Nonetheless, in thiscase the contribution of useful and interferingsignals toward the satisfaction of any RF-EHrequirements is equally important. This implica-tion changes completely the design philosophy ofsuch networks, as interference becomes useful.

This concept was demonstrated for the multi-ple-input single-output (MISO) interferencechannel where K transmitters, each with K anten-nas, communicate with K single-antennareceivers [13]; each receiver is characterized byboth quality-of-service (QoS) and RF-EH con-straints, while PS is used for simultaneous infor-mation/energy transfer. The QoS constraintrequires the SINR to be higher than a giventhreshold, while the RF-EH constraint requiresthe power input to the RF-EH circuitry to beabove a threshold. In this framework, an inter-esting non-convex optimization problem arises inselecting the beamforming weights and thepower of the transmitters as well as the powersplitting ratios at the receivers so as to minimizethe total transmit power. The problem can besolved optimally using semidefinite program-ming, while traditional beamformers can beemployed to obtain suboptimal but low-complex-

ity solutions. An interesting conclusion is thatfor ZF beamforming there always exists aunique, optimal, closed-form power allocation.

The benefit of exploiting interference in thecontext of RF-EH is illustrated in Fig. 5, whichdepicts the transmit power ratio between ZF andoptimal beamforming for varying SINR andRF-EH thresholds (K = 8). The figure indicatesthat by exploiting interference, the transmitpower can be significantly reduced, especially forlow SINR. The reason is that for low SINR thereis room to increase interference, which is benefi-cial for RF-EH. In contrast, high SINR thresholdsrequire almost full cancellation of interference;hence, the solutions obtained from ZF arealmost optimal. The benefits of interferenceexploitation can also be seen with respect to theRF-EH constraints: when the RF-EH thresholdincreases, the ZF/optimal power ratio increasesbecause the optimal scheme manages interfer-ence better. However, the effect of the SINRconstraint on the transmit power ratio is moresignificant compared to the RF-EH constraint.

INTERFERENCE-AIDEDSECRECY RATE IMPROVEMENT

Due to the growing number of wireless applica-tions, confidentiality and secret transmission hasbecome an increasingly important issue. Recent-ly, securing wireless communications at the phys-ical (PHY) layer has been studied as acomplimentary measure to upper layer crypto-graphic techniques. In the presence of eaves-droppers who passively overhear the communi-cations, intentional interference plays a key roleto improve the secrecy rate. This is understand-able since interference will affect both systems;however, if properly designed, it can be an advan-tage for the legitimate system. This is indeedtrue, as shown in [14] that the exploration ofaggregated interference together with locationand channel quality information, can significant-ly improve network secrecy. In the following, wereview several approaches that utilize interfer-ence to confuse the eavesdropper in a simplepoint-to-point network.

Consider a basic three-node system that con-sists of a source S, a destination D, and an eaves-dropper E. When S has multiple antennas, it cantransmit an information-bearing signal to D inthe range space of the channel to D and alsogenerate artificial noise (AN) to E in its nullspace simultaneously. In this way, even withoutknowledge of the instantaneous CSI of the eaves-dropper, the generated AN does not interferewith the legitimate receiver D and only affectsthe eavesdropper node E. The same principleapplies if there are trusted helper relays whocould form distributed beamforming to transmitcooperative jamming signals to E.

When neither multiple antennas at S nortrusted helpers are available, the system mustrely on itself to achieve secure communications.To this end, a self-protection scheme has beenproposed that adopts full-duplex (FD) operationat D to improve the secrecy rate [15], as shown inFig. 6. More specifically, an FD receiver is intro-duced that simultaneously receives its data while

Figure 4. Required SNR per transmit antenna for an uncoded SER of 102with increasing numbers of users and transmit antennas.

Number of users = number of Tx antennas42

15

SNR

per

Tx a

nten

na f

or u

ncod

ed S

ER=

10-2

25

30

35

40

6 8 10 12 14 16

Zero forcing precodingInterference exploitation precoding

8PSK

QPSK

20

ZHENG_LAYOUT.qxp_Layout 10/29/14 3:47 PM Page 156

IEEE Communications Magazine November 2014 157

transmitting a jamming signal to confuse E. Theproposed approach uses intentional interferenceat D to confuse the eavesdroppers and does notrequire external helpers or data retransmission.Due to the FD operation, the receiver experi-ences a loop interference (LI) introduced by thetransmitted jamming signal. If D has multipletransmit or receive antennas, it can employ jointtransmit and receive beamforming for simultane-ous signal detection, suppression, and intentionaljamming.

In Fig. 6 the achievable secrecy rate is evalu-ated against the transmit SNR. We simulate twocases: single transmit/receive-antenna receiverand eavesdropper; and the receiver has twotransmit and two receive antennas while theeavesdropper has four antennas for fairness. Forthe single-antenna case, it is seen that the FDscheme outperforms the half-duplex (HD) oper-ation for transmit SNR greater than 10 dB, anddouble secrecy rate is achieved in the high SNRregion. The performance of the HD scheme sat-urates when the transmit SNR is higher than 20dB. When the receiver has multiple antennasand the eavesdropper adopts a simple maximal-ratio combining (MRC) receiver, the secrecyrate strictly increases with the transmit SNR anddoes not saturate at high SNR as in the HDcase. When the eavesdropper is aware of the FDoperation at D and adopts the minimum-mean-square-error (MMSE) receiver to mitigate thejamming signals from D, the achievable secrecyrate saturates at a high SNR of 40 dB but is stillsignificantly higher than the case with HD receiv-er. This reveals the great potential of usinginterference at the receiver side to provide self-protection against eavesdropping.

CONCLUSIONSIn this article we have introduced radical viewson interference in wireless networks. Traditionalinterference mitigation techniques are no longeroptimal, and innovative ways of utilizing interfer-ence are emerging. As more aggressive resourcesharing and tighter cooperation are foreseen infuture wireless networks, interference manage-ment will continue to be a growing challenge.

Accordingly, it is essential to further thesenew perspectives on interference for more effi-cient radio resource utilization in advanced wire-less concepts such as large-scale antenna arrays(massive MIMO), multicell cooperation, cogni-tive radio, and heterogeneous networks. Indeed,the employment of massive MIMO in future net-works allows the mitigation of interference usingsimple linear operations. In this way, interfer-ence could be available in the network forother purposes without affecting its perfor-mance; this scenario motivates new services andapplications. In future cloud radio access net-works, baseband processing will be shifted fromthe BSs to the central baseband unit pool tojointly process data to and from multicells, andthis offers great opportunities to fully utilizeinterference. In cognitive radio, the interferencefrom the secondary user to the primary user canfacilitate RF energy transfer and be tuned intouseful signals if the primary data are known at

Figure 5. Transmitted power benefit from optimal exploitation of interfer-ence compared to ZF beamforming to achieve SINR and RF-EH con-straints in the MISO interference channel.

SINRth [dB] RF-EH t

h [d

Bm]

10

10

1

20

30

40

50

1520

2530

35 -30-25

5

-20-15

-10-5

0

5

10

15

20

25

30

35

40

tota

lop

t/ P

tota

lZF

P

Figure 6. Left: FD operation at the receiver that creates self-interference to improve the secrecy rate;right: achievable secrecy rate in bits per channel use (bpcu) vs transmit SNR in dB.

Total transmit SNR (dB)50

2

0

Secr

ecy

rate

(bp

cu)

4

6

8

10

12

10 15 20 25 30 35 40

E

DS

HD, multiple-antennaFD, MMSE receiver at D, MRC receiver at EFD, MMSE receiver at D, MMSE receiver at EHD, single-antennaFD, single-antenna

ZHENG_LAYOUT.qxp_Layout 10/29/14 3:47 PM Page 157

IEEE Communications Magazine November 2014158

the secondary user. Regarding security in het-erogeneous networks, a promising direction is tostudy how network interference can be engi-neered to best benefit wireless network secrecy.

ACKNOWLEDGMENTThis work was partially supported by theResearch Promotion Foundation, Cyprus, underthe project KOYLTOYRA/BP-NE/0613/04Full-Duplex Radio: Modeling, Analysis andDesign (FD-RD).

REFERENCES[1] M. Costa, Writing on Dirty Paper, IEEE Trans. Inf. The-

ory, vol. IT-29, May 1983, pp. 43941.[2] A. El Gamal and Y. H. Kim, Network Information Theo-

ry, Cambridge University Press, 2012.[3] B. Nazer and M. Gastpar, Reliable Physical Layer Net-

work Coding, Proc. IEEE, vol. 99, Mar. 2011, pp.43860.

[4] S. A. Jafar, Interference Alignment: A New Look at Sig-nal Dimensions in a Communication Network, Founda-tions and Trends in Communications and InformationTheory, vol. 7, Issue 14, 2010, pp. 113.

[5] S. A. Jafar, Topological Interference ManagementThrough Index Coding, IEEE Trans. Inf. Theory, vol. 60,no. 1, Jan. 2014, pp. 52968.

[6] J. Lee, D. Toumpakaris, and W. Yu, Interference Miti-gation via Joint Detection, IEEE JSAC, vol. 29, no. 6,Jun. 2011, pp. 117284.

[7] C. Masouros, Correlation Rotation Linear Precoding forMIMO Broadcast Channel, IEEE Trans. Sig. Process.,vol. 59, no. 1, pp. 252-262, Jan. 2011.

[8] C. Masouros et al., Known Interference in WirelessCommunications: A Limiting factor or a PotentialSource of Green Signal Power? IEEE Commun. Mag.,vol. 51, no. 10, Oct. 2013, pp.16271.

[9] R. T. Krishnamachari and M. K. Varanasi, InterferenceAlignment Under Limited Feedback for MIMO Interfer-ence Channels, IEEE Trans. Sig. Process., vol. 61, no.15, Aug. 2013, pp. 390817.

[10] N. Lee, J.-B. Lim, and J. Chun, Degrees of Freedom ofthe MIMO Y Channel: Signal Space Alignment for Net-work Coding, IEEE Trans. Inf. Theory, vol. 56, no. 7,Jul. 2010, pp.333242.

[11] Z. Ding and H. V. Poor, A General Framework of Pre-coding Design for Multiple Two-Way Relaying Commu-nications, IEEE Trans. Sig. Process., vol. 61, no. 6, Mar.2013, pp.153135.

[12] R. Zhang and C. K. Ho, MIMO Broadcasting forSimultaneous Wireless Information and Power Trans-fer, IEEE Trans. Wireless Commun., vol. 12, no. 5, May2013, pp. 19892001.

[13] S. Timotheou et al., Beamforming for MISO Interfer-ence Channels with QoS and RF Energy Transfer, IEEETrans. Wireless Commun., vol. 13. no. 5, May 2014, pp.264658.

[14] A. Conti et al., Interference Engineering for Heteroge-neous Wireless Networks with Secrecy, Proc. AsilomarConf. Signals, Systems, and Computers, Pacific Grove,CA, Nov. 2013, pp. 30812.

[15] G. Zheng et al., Improving Physical Layer SecrecyUsing Full-Duplex Jamming Receivers, IEEE Trans. Sig.Process., vol. 61, no. 20, Oct. 2013, pp. 496274.

BIOGRAPHIESGAN ZHENG is currently a Lecturer at the School of Com-puter Science and Electronic Engineering, University ofEssex, UK. He received a B. E. and an M. E. from TianjinUniversity, Tianjin, China, and a Ph.D. degree in electricaland electronic engineering from The University of HongKong, Hong Kong, in 2008. He worked as a research asso-

ciate at University College London, UK, and the Universityof Luxembourg, Luxembourg. His research interestsinclude cooperative communications, cognitive radio,physical-layer security, and full-duplex radio. He was thefirst recipient of IEEE Signal Processing Letters Best PaperAward in 2013.

IOANNIS KRIKIDIS received a diploma in computer engineeringfrom the Computer Engineering and Informatics Depart-ment (CEID) of the University of Patras, Greece, in 2000,and the M.Sc. and Ph.D. degrees from Ecole NationaleSuprieure des Tlcommunications (ENST), Paris, France,in 2001 and 2005, respectively, all in electrical engineering.From 2006 to 2007 he worked as a post-doctoralresearcher, with ENST, Paris, France, and from 2007 to2010 he was a research fellow at the School of Engineer-ing and Electronics at the University of Edinburgh, Edin-burgh, UK. He is currently an assistant professor in theDepartment of Electrical and Computer Engineering, Uni-versity of Cyprus, Nicosia, Cyprus. His current researchinterests include information theory, wireless communica-tions, cooperative communications, cognitive radio, andsecrecy communications.

CHRISTOS MASOUROS is currently a lecturer in the Dept. ofElectrical & Electronic Eng., University College London. Hereceived his diploma in electrical and computer engineer-ing from the University of Patras, Greece, in 2004, an MSc.by research and a Ph.D. in electrical and electronic engi-neering from the University of Manchester, UK in 2006 and2009, respectively. He has previously held a research asso-ciate position at the University of Manchester, UK, and aresearch fellow position in Queens University Belfast, UK.He holds a Royal Academy of Engineering Research Fellow-ship 2011-2016. His research interests lie in the field ofwireless communications and signal processing with partic-ular focus on green communications, large scale antennasystems, cognitive radio, interference mitigation techniquesfor MIMO, and multicarrier communications.

STELIOS TIMOTHEOU holds a Dipl.-Ing from the Electrical andComputer Engineering School of the National TechnicalUniversity of Athens, and an M.Sc. and Ph.D. from theElectrical and Electronic Engineering Department of Imperi-al College London. He is currently a research associate atthe KIOS Research Center for Intelligent Systems and Net-works of the University of Cyprus. His research focuses onthe modeling and solution of problems that arise in com-munication systems, intelligent transportation systems anddisaster management by developing optimization, machinelearning and computational intelligence techniques.

DIMITRIS TOUMPAKARIS is an assistant professor in the Depart-ment of Electrical and Computer Engineering at the Univer-sity of Patras, Greece. He holds a diploma in electrical andcomputer engineering from the National Technical Universi-ty of Athens, Greece, and a M.S. and Ph.D. in electricalengineering from Stanford University. His current researchinterests include baseband system design, interferencemanagement, and multiuser information theory. He hasbeen an editor for IEEE Communications Letters since2012.

ZHIGUO DING is a chair professor at the School of Comput-ing and Communications, Lancaster University, UK. Hisresearch interests are game theory, cooperative and energyharvesting networks, and statistical signal processing. Hewas co-chair of the WCNC-2013 Workshop on NewAdvances for Physical Layer Network Coding, and is servingas an editor for IEEE Transaction on Communications, IEEEWireless Communications Letters, IEEE Communication Let-ters, and the Journal of Wireless Communications andMobile Computing. He received the best paper award atthe IET Comm. Conf. on Wireless, Mobile and Computingin 2009, was named by IEEE Communications Letters as anExemplary Reviewer in 2012, and was awarded the EUMarie Curie Fellowship for 2012-2014.

Traditional interfer-

ence mitigation tech-

niques are no longer

optimal, and innova-

tive ways of utilizing

interference are

emerging. As more

aggressive resource

sharing and tighter

cooperation are fore-

seen in future wire-

less networks,

interference

management will

continue to be a

growing challenge.

ZHENG_LAYOUT.qxp_Layout 10/29/14 3:47 PM Page 158

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages false /GrayImageMinResolution 300 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages false /GrayImageDownsampleType /Average /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages false /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages false /MonoImageDownsampleType /Average /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier (CGATS TR 001) /PDFXOutputCondition () /PDFXRegistryName (http://www.color.org) /PDFXTrapped /False

/CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles true /MarksOffset 6 /MarksWeight 0.250000 /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /UseName /PageMarksFile /RomanDefault /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /LeaveUntagged /UseDocumentBleed false >> > ]>> setdistillerparams> setpagedevice