Securing Underwater Acoustic Communications Through Analog Network Coding
Securing Underwater Acoustic Communicationsthrough Analog Network CodingHovannes Kulhandjian†, Tommaso Melodia† , and Dimitrios Koutsonikolas‡†Department of Electrical EngineeringDepartment of Computer Science and EngineeringState University of New York at Buffalo, Buffalo, New York 14260E-mail: {hkk2, tmelodia, dimitrio}@buffalo.edu‡Abstract—We propose a new secure underwater acoustic communication scheme designed to let a user (Alice) transmit aconfidential message to another user (Bob) in the presence of aneavesdropper (Eve). A typical approach in conventional wirelessphysical-layer security is to rely on a friendly jammer to jamEve through artificial noise (AN). Instead, for the first time,we propose a secure underwater communication scheme thatrelies on cooperative friendly jamming through CDMA-basedanalog network coding (ANC). The cooperative friendly jammertransmits information using the same spreading code used inthe legitimate Alice-Bob link. The information transmitted bythe cooperative jammer is known a priori to Bob, but not toEve. Although the jammer’s packet will also interfere at Bob,we show that after jointly estimating the two multipath-affectedchannels, Bob can suppress the interfering packet and decodeAlice’s packet, while Eve cannot. We also formulate the problemof joint optimal selection of friendly jammer and power allocation(for Alice and the jammer) that minimize Eve’s capability ofintercepting the signal while guaranteeing a predefined levelof quality of service (QoS) for Bob. The proposed scheme isimplemented in a testbed based on Teledyne Benthos TelesonarSM-975 underwater modems and tested extensively in LakeLaSalle at the University at Buffalo. Experiments and simulationsdemonstrate that, for a given energy budget, the proposed schemecan guarantee much higher bit error rate (BER) at Eve, whilecreating minimal BER disturbance at Bob, compared to the ANaided approach.I. I NTRODUCTIONUnderwater acoustic communications and networking haverecently attracted considerable attention due to increasinginterest in many undersea commercial and military applications [1], [2], [3], [4]. Although radio frequency (RF)electromagnetic and optical waves are the dominant physicalcommunication carriers in terrestrial wireless communications,in water they are severely affected by high attenuation andscattering, respectively. Acoustic communication is thereforethe transmission technology of choice for wireless underwaternetworked systems [1].The underwater acoustic (UW-A) channel is considered oneof the most challenging environments to establish reliableand secure communications. Some of the challenges includeslow propagation of acoustic waves, limited bandwidth, andhigh and variable propagation delays. Furthermore, the UW-Achannel is affected by Doppler spread and by severe timevarying multipath fading [1], [2].Such a challenging environment makes reliable communications hard to achieve, and at the same time makes underwaterAcknowledgment: This work was partially supported by the NationalScience Foundation under grants CNS-1055945 and CNS-1126357.978-1-4799-4657-0/14/ 31.00 c 2014 IEEEnetworks prone to malicious attacks. Some security challengesin underwater networks are discussed in [5]. In this paper,we concentrate on the problem of transmitting securely aconfidential message in the presence of eavesdropping attacks.One way to overcome eavesdropping is to apply cryptographicapproaches at the upper layers of the protocol stack byencrypting data before transmission. However, cryptographicmechanisms can face potential attacks at the higher layers [6],and suffer from heavy computational complexity, especially,in resource constrained underwater acoustic sensor networks(UW-ASNs) [1], [3]. In any case, it is desirable to improve thesecurity of the physical layer wireless channel by impairingthe eavesdroppers’ intercepting capabilities in the first place[7].Physical layer security has therefore recently attracted considerable attention [8], [9], [10] due to its inherent ability toprevent eavesdropping. Although most research has focusedon information theoretic approaches [11], [12], [13], the topichas drawn significant interest from the signal processing [6],[7], [14] and networking [15], [16] communities. However,very limited research only has addressed secure UW-A communications in the presence of an eavesdropper.Among these, in [17], a direct-sequence spread-spectrum(DSSS) waveform design with low probability of interception(LPI) was proposed to provide covert UW-A communications.Similarly, in [18], a multi-carrier spread-spectrum (MCSS)modulation was proposed as a means to render covert UW-Acommunication at low signal-to-noise ratio (SNR). A receiverwith multiband equalization was proposed to jointly equalize and despread the contiguous frequency bands carryingthe same symbol stream. In [19], a multiband orthogonalfrequency-division multiplexing (OFDM) transmitter and receiver were presented for secure UW-A communications atlow SNR regime with the intention to avoid interception.However, those schemes may become vulnerable to eavesdropping if the adversary is able to identify the spreading codeor the modulation technique used by the two parties. Directsequence code-division multiple-access (DS-CDMA) schemeshave long been used to provide covert communications [20],[21]. However, recent work [22] has shown that DS-CDMAcan become vulnerable to attacks since it is possible to blindlyidentify the spreading code used by the legitimate user whenneither channel state information nor training sequence isavailable. Accordingly, it is necessary to explore alternativemeans to provide security at the physical layer.In this paper, we propose a new secure UW-A commu-
nication scheme designed to let a user (Alice) transmit aconfidential message to another user (Bob) in the presence ofan eavesdropper (Eve). In the case when the adversary has abetter channel quality, compared to the legitimate link, perfectsecrecy (i.e., zero information leakage to the eavesdropper bylistening to the source-destination message exchange) can notbe achieved. To overcome this obstacle, a cooperative friendlyjammer is often introduced to degrade the adversary’s channel[6], [14]. A common approach frequently used by cooperativefriendly jammers is to jam the eavesdropper through artificialnoise (AN) [23], [24]. Since such an approach can also degradethe channel of the legitimate user, oftentimes, an array beamforming approach using multiple antennas is utilized to designa scheme such that most of the AN jamming signal is targetedto the adversary’s location, while minimizing its effects at theintended user [23], [24]. Usually, a perfect knowledge of theeavesdropper’s channel condition is necessary to design suchschemes [23], which may be hard to obtain or not be availablealtogether. Moreover, in the case when the adversary is in closeproximity of the legitimate user, even a beamforming approachcannot be of much help to avoid degrading the channel of thelegitimate user. Besides, beamforming requires nodes to beequipped with arrays of transducers, which can be very costlyto provide in underwater sensor network deployments [1].Therefore, for the first time, we propose a secure underwatercommunication scheme that, unlike previous work relying onAN-based jamming, is based on cooperative friendly jammingbuilt upon CDMA-based analog network coding (ANC), atechnique developed in our recent work [25]. The basic ideaof ANC [26], also known as physical layer network coding(PNC) [27], is to allow concurrent transmissions of signalsover the wireless medium so that they intentionally interferewith each other. The receiver, having heard the interferedsignal from prior transmissions, will suppress the interferencebefore decoding the desired information [25]. Prior work hasused ANC as a technique to increase the network throughput.To the best of our knowledge, our work is the first to use theprinciple of ANC with a completely different objective, i.e.,to provide covert communications in UW-A channels.The CDMA-based ANC scheme differs from conventionalDS-CDMA interference cancellation in terms of the natureof the interference signal that each scheme is designed tosuppress. In conventional DS-CDMA, multiple access interference (MAI) is generated by different users utilizing uniquespreading codes. Instead, in CDMA-based ANC, MAI isgenerated by a pair of nodes accessing the channel usingthe same spreading code. As a consequence, it is much morechallenging to cancel the resulting interference.The core idea of our proposed J-ANC (Jamming-throughANC) scheme is as follows. We consider a DS-CDMA link between Alice and Bob1 . Eve may be located closer to Alice thanBob, and thus may have a better signal/channel quality relativeto Bob. To prevent Eve from intercepting Alice’s packet, acooperative friendly jammer is selected to transmit informationmodulated using the same spreading code assigned to thelegitimate Alice-Bob link. Although we could let Alice mix1 CDMA is one of the most promising physical layer and multiple accesstechniques for UW-ASNs [1], since it is robust to frequency-selective fadingand can compensate for the effect of multipath through RAKE receivers [4].the jamming signal in the digital domain (i.e., using networkcoding [28]) before transmission, by introducing a cooperativefriendly jammer we leverage the physical properties of thewireless medium and thus make it even harder for Eve tointercept the communication, since she will have to jointlyestimate the two channels and remove the jamming signalbefore being able to retrieve Alice’s packet. The informationbits transmitted by the cooperative jammer are randomlygenerated and are known a priori to Bob, but not to Eve.Although the jammer’s packet will also interfere at Bob, weshow that after jointly estimating the two multipath affectedchannels, Bob can suppress the interfering signal and retrieveAlice’s packet. Therefore, Bob will be able to decode Alice’spacket, while Eve will fail to do so with high probability.We also formulate the problem of optimal selection of thefriendly jammer among a set of jammers and optimal energyallocation for both Alice and the jammer, with the objectiveto guarantee a minimum level of signal-to-interference-plusnoise ratio (SINR) to Bob and, at the same time, degrade theSINR of Eve as much as possible.The contributions of this paper are outlined as follows.1) For the first time, we propose a secure underwater communication scheme that, unlike previous work relying on ANbased jamming, is based on cooperative friendly jamming builtupon CDMA-based ANC. To prevent Eve from interceptingAlice’s packet, a cooperative friendly jammer is selected totransmit information modulated using the same spreading codeassigned to the legitimate Alice-Bob link. The informationtransmitted by the cooperative jammer is known a priorito Bob, but not to the eavesdropper.2) To the best of our knowledge, the case when Eve is invery close proximity of the legitimate user is not addressedadequately by the physical layer security research community.This can be justified by the fact that perfect secrecy cannotbe achieved. The proposed scheme can even provide securitywhen Eve is located nearby Bob such that she can overhearwhat Bob receives but will not be able to decode the message.3) We formulate the problem of optimal selection of thefriendly jammer among a set of jammers and optimal energyallocation for both Alice and the jammer, with the objective toguarantee a minimum level of SINR to Bob and at the sametime, degrade the SINR of Eve as much as possible.4) We implement J-ANC in a testbed based on TeledyneBenthos Telesonar SM-975 underwater modems and test itextensively in Lake LaSalle at the University at Buffalo.Experiments and simulations demonstrate that J-ANC is lessharmful to the intended receiver and can provide higher security against an eavesdropper compared to traditional AN-aidedapproaches. Specifically, we show that for a given jammingenergy budget, J-ANC can guarantee much higher BER at Eve,while causing minimal BER disturbance at Bob, compared tothe AN-aided approach.The rest of this paper is organized as follows. In SectionII, we describe the system model with detailed discussions onjoint channel estimation and receiver design for both Bob andEve. In Section III, we address the friendly jammer selectionwith optimal energy allocation problem. In Section IV, weevaluate the proposed scheme. Finally, in Section V, we drawthe main conclusions.
Notation: The following notation is used throughout thepaper. Boldface lower-case letters indicate column vectors,boldface upper-case letters indicate matrices, xH denotes theHermitian of vector x, IN and 0N are the identity and zeromatrices of dimensions N N respectively, tr{X} representsthe trace of a matrix X, C is the set of all complex numbers,E {·} represents statistical expectation, · and k · k are themagnitude and the norm of a scalar and vector, respectively,Re(·) denotes the real part of a complex valued vector andsgn(·) denotes zero-threshold quantization.II. S YSTEM M ODEL AND P ROBLEM F ORMULATIONWe consider a CDMA-based network shown in Fig. 1, inwhich Alice would like to transmit confidential information toBob in the presence of an eavesdropper, Eve. We assume thatEve can be located closer to Alice (thus she may have a bettersignal/channel quality) compared to Bob, as depicted in Fig. 1.To prevent Eve from intercepting Alice’s packet, we utilize afriendly jammer. Using the same DS-CDMA spreading code,Alice and the jammer concurrently transmit packets, PA andPJ , containing independent and identically distributed (i.i.d.)bit streams, to Bob and Eve, respectively. Due to the broadcastnature of the wireless medium, both Bob and Eve will overhearthe packets, PA and PJ . We further assume that Eve hasperfect knowledge of all the channel state information (CSI)(between Alice/jammer and itself) and of the spreading codeutilized by Alice/jammer. In other words, we consider theworst case for Bob, but best case for Eve. The packet PJis assumed to be known to both Alice and Bob but not toEve. We now illustrate a technique through which Bob cansuccessfully retrieve Alice’s packet at high SNR, while Evewill fail to do so with high probability, which is our ultimategoal.equiprobable random variables that are independent across theusers (k A, J) and within a user steam (i 1, 2, .) [22].Before transmission of the information bits, each chip of thespreading sequence is multiplied by a pulse shaping signal p(t)and a carrier. The normalized modulated spreading sequencein the time domain is then denoted byLs(t) 1Xd(l)p(t lTc)ej2πfc t ,L(3)l 1where d(l) { 1, 1} is the lth transmitted chip of thespreading sequence, Tc T /L is the chip period, T is theinformation bit duration, and fc is the is the carrier frequency.We assume that packets propagate over multipath Rayleighfading UW-A channels, which is commonly used in modelingUW-A channels [29], [30]. Without loss of generality, and forease of exposition, we assume here that packets transmittedby Alice and the jammer experience the same number ofresolvable multipaths, MAB MJB M , and arrive atBob and Eve simultaneously. However, we note that theproposed scheme does not require synchronous arrival of thetwo packets at the relay2 , which is difficult to achieve due tolong propagation delays in UW-A channel3 . A preamble and apostamble, identical chirp signals of duration 100ms sweepingthe bandwidth from 10Hz to 2.6kHz appended to each packet,are used for channel probing, symbol synchronization for chipmatched filtering, and multipath delay spread estimation, aswill be further discussed in Section IV-B.After carrier demodulation, chip-matched filtering and sampling at the chip rate over a multipath extended bit periodof L M 1 chips, where M is the number of resolvable multipaths, the received signals, rB (i) CL M 1 andrE (i) CL M 1 (that is, the noise-affected superimposedversion of the ith bits of Alice and the jammer) at Bob andEve, respectively, are denoted bypprB (i) EA SA (i)hAB EJ SJ (i)hJB nB (i), (4)pprE (i) EA SA (i)hAE EJ SJ (i)hJE nE (i), (5)whereFig. 1.System model.Because of the highly frequency-selective distortion causedby multipath propagation in the UW-A channel, it is essentialto estimate the CSI periodically [25]. Accordingly, we utilizea set of Np pilot bits that are repeated periodically and areinserted in each packet distanced less than the coherence time,TCT , of the channel. The pilot bits will be used to jointlyestimate the CSIs from Alice-to-Bob and Jammer-to-Bob.The baseband supervised and information bits transmittedby Alice and the friendly jammer arepi 1, 2, . ,(1)xA (i) bA (i) EA s,pxJ (i) bJ (i) EJ s,i 1, 2, . ,(2)where s L1 { 1}L denotes the normalized spreading codeof length L used by Alice and the jammer, EA and EJdenote the transmit energy per bit, bA (i), bJ (i) { 1, 1}are ith pilot/information bits binary phase-shift-keying (BPSK)data modulated and transmitted by Alice and the jammer,respectively. The information bits bk (i) are viewed as binaryhAB [hAB (1), hAB (2), . . . , hAB (M )]H ,(6)H(7)HhJB [hJB (1), hJB (2), . . . , hJB (M )] ,(8)hJE [hJE (1), hJE (2), . . . , hJE (M )]H ,(9)hAE [hAE (1), hAE (2), . . . , hAE (M )] ,are multipath channel coefficients from Alice-to-Bob, Aliceto-Eve, Jammer-to-Bob and Jammer-to-Eve of lengths M ,respectively. hAB (q), hAE (q), hJB (q) and hJE (q) representthe q th resolvable path coefficients modeled as quasi-staticRayleigh-distributed random variables that remain constantduring TCT block length, nB and nE are ambient noise, and SA (i) , S0A (i) S A (i) SA (i),SJ (i) ,S0J (i) S J (i) S J (i),(10)(11)2 Due to space constraints, we will address the asynchronous case withdifferent number of resolvable multipath arrivals at the relay at a later time.3 In practice, the packet transmitted by the jammer, P , is slightly longerJthan Alice’s packet, to make sure that all the bits of PA are jammed at Eve.
wheres(1) 0 .s(1) . . s(L) .S0k (i) , bk (i) 0 s(L) . .00 0. . 0 S k (i) , bk (i 1) s(1) . . . 0. . 0 ,(12) s(1) . .s(L) (L M 1) M0 0 . . . 0 0 .,(13). 0 .0 . .s(M 1) . . . s(1) 0(L M 1) M 0 s(L) . . . s(L M 1). . . . 0 . 0 . . s(L)Sk (i) , bk (i 1) (14) . 0 0 0 . .0 0 .0(L M 1) M The matrices S0k (i), S k (i) and Sk (i) correspond to thespreading code matrices generated due to the transmission ofbits b(i), b(i 1) and b(i 1), respectively, by user k {A, J},Alice or the cooperative jammer in this case.Joint Channel Estimation at Bob. We now show how Bobcan estimate the CSIs from Alice-to-Bob and Jammer-to-Bobjointly. Let us defineihppEA SA (i), EJ SJ (i) ,(15)SAJ (i) ,We can obtain an accurate estimate of hAB,JB if and only if(15) is of full rank. This condition is satisfied if (15) contains atleast 2M independent vectors. In this work, we use columnsof a Sylvester-Hadamard matrix, HL with elements 1 or 1, of order L 2n , n 2, 3, . , as our spreading code.The Sylvester-Hadamard matrix has good autocorrelation andcross-correlation properties [31]. Rows (and columns) of theHL are mutually orthogonal to each other. For a spreadingcode of order L 4 extracted from HL , the above conditioncannot be satisfied for M 3, hence a spreading code lengthof L 8 or longer needs to be used in this case. In practice,as we will show in Section IV, a spreading code of length atleast L 32 might be necessary in shallow UW-A channels.In addition to that, it is important to select the trainingsequences for both nodes with very low cross-correlationproperties to minimize the noise enhancement,S†AJ (i)nB (i),n 1 owhich can be shown to be equal to tr SAJ (i)H SAJ (i)[25]. Accordingly, to minimize the noise enhancement, weutilize two orthogonal sequences of pilot bits extracted againfrom columns of HL , of order L 2n , n 4, 5, . .The CSIs from Alice-to-Bob and Jammer-to-Bob are thencomputed asĥAB [IM 0M ] ĥAB,JB ,(21)ĥJB [0M IM ] ĥAB,JB .Receiver Design at Bob. To decode the information bits, Bobwill use the estimated CSIs, ĥAB , ĥJB , and design a RAKEmatched-filter that decides on the transmitted bit of the userof interest (Alice) based on the sum of the individual M pathcorrelator outputs; which can be equivalently characterized bythe normalized static (L M 1)-tap FIR filter given bywMFB (L M 1) 2MhAB,JB ,"#hAB.hJB(16)wherei 1, 2, . .(17)Before jointly estimating the channel coefficients, hAB andhJB , we first define the pseudo-inverse of SAJ (i) for (L M 1) 2M using the Moore-Penrose pseudo-inverse formula as 1S†AJ (i) , SAJ (i)H SAJ (i)SAJ (i)H .(18)If we assume that nB is modeled as white Gaussian distributed, the conditional maximum-likelihood (ML) estimateof hAB,JB , for a given supervised bit i, can be obtained byminimizing the following squared error quantityh̃AB,JB argminhAB,JB C2MkrB (i) SAJ (i)hAB,JB k22 . (19)Since the channel noise is assumed to follow a zero-meanGaussian distribution, the marginal solution of (19) can beestimated by sample averaging over a data record of Np pilotbits asNp1 X †S (i) rB (i).(20)ĥAB,JB Np i 1 AJSMF ĥAB,HĥAB SHMF SMF ĥAB s(1) 0 . . . 0 . .s(1). . . 0 s(L) . , 0 s(L)s(1) . . . 00 . . . s(L) (L M 1) M(23) 2M 1We rewrite (4) in a more compact form asrB (i) SAJ (i)hAB,JB nB (i),(22)SMF(24)represents the M path-correlator outputs, which can be constructed knowing the number of multipaths, M , which isestimated through chirp-matched filtering (Section IV-B).Before decoding the information bits, we first cancel theinter-symbol-interference (ISI) resulting from the previouslydecoded bits of Alice. Moreover, we cancel the estimatedinterfered data bits from the jammer and, employing theRAKE-matched-filter proposed in (23), the information bits(j 1, 2, .) of the user of interest (Alice) are decoded as i h H,r̂(j) S(j)ĥ S(j)ĥb̂A (j) sgn Re wMFBABJJBAB(25)where S A (j)ĥAB is the ISI of the previously decoded bit fromAlice and SJ (j)ĥJB is the estimate of the interfered data bit
originally transmitted by the jammer.Receiver Design at Eve. We consider the worst case scenario,that is, Eve knows all the CSIs, the number of multipathM , and the spreading code used by Alice/jammer. Beforeattempting to decode Alice’s packet, PA , Eve will first designa linear maximum SINR filter ason 2H( EA SA (i)hAE )E wEowmaxSINRE arg min n 2wEH( E S (i)hE wEJ JJE nE (i)) kR 1E SMF hAE ,H 1 EA hHAE SA RI NE SA hAE ,III. JAMMER S ELECTION WITH O PTIMAL E NERGYA LLOCATIONIn the case where a set of J cooperative friendly jammers(i.e., J1 , J2 , ., JJ ) are available, selecting the best jammerand optimally allocating constrained energy resource to bothAlice and the jammer may further enhance the performance ofour proposed J-ANC scheme. Consider the network topologyshown in Fig. 1. We assume that Alice is in the vicinity of aset of J friendly jammers that are willing to cooperate withher to jam Eve. We address two cases, i.e., (i) known channelconditions of Eve, and (ii) unknown channel conditions of Eve.Note that Alice may infer the CSI of Eve in cases where itknows the positions of Eve and of the friendly jammer [7]. Forexample, when Eve is located close to Bob, after estimatingthe CSIs, Bob can send them to Alice. The case of unknownCSI of Eve is likely to be the more common case.i) Known channel conditions of Eve. We assume that Aliceknows the CSIs, Jammer-to-Bob and Jammer-to-Eve, J J .As we will show, Alice will pick the jammer that will domaximum harm to Eve, while minimizing its jamming effecton the intended receiver, Bob.Our objective is to jointly select (a) the optimal transmitenergies (per bit), EA and EJ , and (b) the jammer, to minimizethe SINR of Eve4 while guaranteeing a minimum level ofSINR (QoS) (which can be translated to BER) to Bob.We formulate the optimization problem asminimize SIN RE(28)subject to : SIN RB β,0 EA Emax ,0 EJ Emax ,(29)(30)(31)where β is the minimum SINR (QoS) requirement of Bob,4 Ideally, we would like to have Eve’s BER be as close as possible to 1/2.When her BER is 1/2, or equivalently her SINR is 0dB, the transmissioncan be called perfectly secure [7].(32)andSIN RB ,(26)where k 0, and RE is the autocorrelation matrix of theobserved signal at Eve given by (5) (which can be estimatedby sample averaging).Eve will attempt to decode the information bits transmittedby Alice using the linear maximum SINR filter (26), Hr (j) , j 1, 2, . . (27)b̃A (j) sgn Re wmaxSINRE EEA ,EJ ,J Jon 2HESh)E wmaxSINR(AAAEEoSIN RE , n 2HE wmaxSINRE( EJ SJ hJE nE)no 2HE wMF(ESh)A A ABB 2 H ( E S hE wMF ESĥ n)JJJBJJJBBBH 1 EA hHAB SA RI NB SA hAB ,(33)are the SINRs perceived at Eve and Bob, respectively, where p p H RI NE EEJ SJ hJE nEEJ SJ hJE nEH2 EJ SJ hJE hHJE SJ σE I(L M 1) ,(34)and n pEJ SJ (hJB ĥJB ) nBRI NB E p H EJ SJ (hJB ĥJB ) nBHH2H EJ (SJ hJB hHJB SJ SJ ĥJB ĥJB SJ ) σB I(L M 1) ,(35)are the autocorrelation matrix of combined interference andnoise at Eve and Bob, respectively. Notice that Bob will beable to suppress most of the interference term ( EJ SJ hJB )caused by the jammer, but Eve will not.It is easy to verify that, at optimality, (29) must hold withequality. Therefore, from (28), (29) and (33) it can be shownthat the minimum energy to be allocated to Alice should satisfyoptβ. Accordingly, we can reformulateEA hH SH R 1S hAB AI NB A ABthe optimization problem (28)-(31) asminimizeH 1βhHAE SA RI N E SA hAEH 1hHAB SA RI N B SA hABβsubject to : H H 1 Emax ,hAB SA RI N B SA hABEJ ,J J0 EJ Emax .(36)(37)(38)By solving the optimization problem (36) - (38), we obtainEJopt Emax . Note that in case the constraint (37) cannot besatisfied for a specific SINR requirement of Bob (e.g. due topoor channel conditions), Alice and the jammer will choosenot to transmit their packets.Assuming the CSIs of Eve (hAE , hJE , J J ) and thedisturbance autocorrelation matrix (34) are available to Alice,the optimal jammer selection problem becomes a combinatorial optimization problem. To find the best jammer, Alicewill compute (36) for each available jammer, and the one thatprovides the lowest value will be selected as the optimumjammer, J opt .ii) Unknown channel conditions of Eve. In case the locationof Eve is unknown or the disturbance autocorrelation matrix aswell as the CSIs of Eve cannot be computed adaptively, Alicewill select a friendly jammer that is closest to her. Having
selected the jammer, Alice and the jammer will use the optimalenergy allocation, discussed above, to jam the eavesdropper.The idea behind selecting the jammer closest to Alice is tolet the two signals (i.e. the jammer’s and hers) attenuate withsimilar amount such that the SINR at Eve with high probabilitywill be less than 0 dB, irrespective of the location of Eve.optNote that the optimal energy allocation, EAand EJopt , donot depend on the CSI of Eve.IV. P ERFORMANCE E VALUATIONWe evaluate the performance of the proposed J-ANCscheme using simulations and underwater testbed experiments.We compare J-ANC against the conventional DS-CDMA,utilizing AN as a jamming source, in terms of average BER.Even though secrecy capacity is a more commonly usedperformance metric in theoretical physical layer security, weconsider BER in our evaluation, as it is more practical andinformative, (i.e., it provides more details on the expectednumber of bits that can be correctly decoded and whether thepacket can successfully be received or not).A. Simulation ResultsSimulations were performed using the system model discussed in Section II. Here, Alice and the friendly jammerconcurrently transmit their packets, using the same spreadingcode, to Bob and Eve, respectively. The channel is modeled asquasi-static frequency-selective Rayleigh fading channels. Themultipath channel coefficients are considered as independentzero-mean complex Gaussian random variables of variance1/M , and the number of multipaths M are randomly selectedfrom the range of values 1–15. Bob jointly estimates theCSIs (Alice-to-Bob and Jammer-to-Bob), and removes theinterference, caused by the jammer’s packet, before decodingthe confidential information transmitted by Alice. AssumingEve has perfect knowledge of CSIs and of the spreading codeutilized by Alice, she will try to intercept Alice’s packet usingthe maximum-SINR filter (26) discussed in Section II.The transmitted packets are divided into fragments and pilotbits are inserted before each fragment. The fragment size isdetermined based on the average coherence time of a shallowwater acoustic channel, which is in the order of a few secondsfor a transmission frequency of 10 kHz [32]. Accordingly,a fragment size of 125 Bytes is selected. We assume Eveis located close to Bob for fair comparison, such that shehas the ability to receive exactly what Bob receives. Unlessotherwise stated, the simulation parameters are payload size 1.25 kBytes, fragment size 125 Bytes, energy per bitEA EJ , number of pilot bits NP 16 and spreadingcode length L 32. The probability of error conditionedon channel coefficients is averaged over 1000 independentchannel realizations.Figure 2 plots the average BER of the proposed J-ANCscheme and conventional DS-CDMA with and without ANjamming for various SNR values of Bob. As we can see, theperformance of J-ANC, in terms of average BER, is closeto the conventional DS-CDMA scheme that does not utilizea cooperative jammer. The price we pay is in joint channelestimation. The better we estimate the channel, the closerthe performance is to the conventional DS-CDMA schemewithout jammer.
in water they are severely affected by high attenuation and scattering, respectively. Acoustic communication is therefore the transmission technology of choice for wireless underwater networked systems [1]. The underwater acoustic (UW-A) channel is considered one of the most challenging environments to establish reliable and secure communications.
underwater communications, with a particular emphasis on acoustic communications. This thesis is the result of a partnership with a team at the Universidade do Minho that has been working in an acoustic underwater communications project. The work presented herein is centered on the channel
Underwater acoustic channels are generally recognized as one of the most difficult communication media in use today. Acoustic propagation is best supported at low frequencies, and the bandwidth available for communication is extremely limited. For example, an acoustic system may operate in a frequency range between 10 and 15 kHz. Although the .
Wireline Quality Underwater Wireless Communication Using High Speed Acoustic Modems Xiaolong Yu, Ph.D. LinkQuest Inc. 7933 Silverton Ave., Suite 717 San Diego, CA 92126 USA Email: xyu@link-quest.com Abstract--The paper introduces LinkQuest Inc.'s cutting-edge high speed underwater acoustic modems. LinkQuest Inc.
communications [41]. Acoustic waves, instead, may enable communications over long-range links since they suffer from relatively low absorption. This has con-tributed to making acoustic transmission the most common underwater communica-tion technique since World War Two [17, 141, 126].
the limited depth of underwater welding. Welding equipment transformed from manual welding to underwater automatic welding. The efficient and low-cost underwater welding was achieved[7]. In order to study the automatic welding technology under larger deep-water environment, the underwater automatic welding system was designed in this paper. The
cast caused by the effects of underwater imaging conditions deteriorate the capability to fully extract valuable information from underwater images for further processing such as marine, mine detection and aquatic robot inspection. Hence, it is of great interest to restore degraded underwater images for high-quality underwater imaging [3].
summarize the associated physical layer challenges for underwater networking. In Sections 5-9 we dis-cuss physical, data link, network, transport and application layer issues in underwater sensor net-works, respectively. In Section 10 we describe some experimental implementations of underwater sensor networks while in Section 11 we draw the
Fjalët kyce : Administrim publik, Demokraci, Qeverisje, Burokraci, Korrupsion. 3 Abstract. Public administration, and as a result all the other institutions that are involved in the spectrum of its concept, is a field of study that are mounted on many debates. First, it is not determined whether the public administration ca be called a discipline in itself, because it is still a heated debate .
