The thesis focuses on the privacy of communication that can be ensured by means of the physical layer, i.e., by appropriately chosen coding and resource allocation schemes. The fundamentals of physical-layer security have been already formulated in the 1970s by Wyner (1975), Csiszár and Körner (1978). But only nowadays we have the technical progress such that these ideas can find their way in current and future communication systems, which has driven the growing interest in this area of research in the last years. We analyze two physical-layer approaches that can ensure the secret transmission of private information in wireless systems in presence of an eavesdropper. One is the direct transmission of the information to the intended receiver, where the transmitter has to simultaneously ensure the reliability and the secrecy of the information. The other is a two-phase approach, where two legitimated users first agree on a common and secret key, which they use afterwards to encrypt the information before it is transmitted. In this case, the secrecy and the reliability of the transmission are managed separately in the two phases. The secrecy of the transmitted messages mainly depends on reliable information or reasonable and justifiable assumptions about the channel to the potential eavesdropper. Perfect state information about the channel to a passive eavesdropper is not a rational assumption. Thus, we introduce a deterministic model for the uncertainty about this channel, which yields a set of possible eavesdropper channels. We consider the optimization of worst-case rates in systems with multi-antenna Gaussian channels for both approaches. We study which transmit strategy can yield a maximum rate if we assume that the eavesdropper can always observe the corresponding worst-case channel that reduces the achievable rate for the secret transmission to a minimum. For both approaches, we show that the resulting max-min problem over the matrices that describe the multi-antenna system can be reduced to an equivalent problem over the eigenvalues of these matrices. We characterize the optimal resource allocation under a sum power constraint over all antennas and derive waterfilling solutions for the corresponding worst-case channel to the eavesdropper for a constraint on the sum of all channel gains. We show that all rates converge to finite limits for high signal-to-noise ratios (SNR), if we do not restrict the number of antennas for the eavesdropper. These limits are characterized by the quotients of the eigenvalues resulting from the Gramian matrices of both channels. For the low-SNR regime, we observe a rate increase that depends only on the differences of these eigenvalues for the direct-transmission approach. For the key generation approach, there exists no dependence from the eavesdropper channel in this regime. The comparison of both approaches shows that the superiority of an approach over the other mainly depends on the SNR and the quality of the eavesdropper channel. The direct-transmission approach is advantageous for low SNR and comparably bad eavesdropper channels, whereas the key generation approach benefits more from high SNR and comparably good eavesdropper channels. All results are discussed in combination with numerous illustrations.
The thesis focuses on the privacy of communication that can be ensured by means of the physical layer, i.e., by appropriately chosen coding and resource allocation schemes. The fundamentals of physical-layer security have been already formulated in the 1970s by Wyner (1975), Csiszár and Körner (1978). But only nowadays we have the technical progress such that these ideas can find their way in current and future communication systems, which has driven the growing interest in this area of research in the last years. We analyze two physical-layer approaches that can ensure the secret transmission of private information in wireless systems in presence of an eavesdropper. One is the direct transmission of the information to the intended receiver, where the transmitter has to simultaneously ensure the reliability and the secrecy of the information. The other is a two-phase approach, where two legitimated users first agree on a common and secret key, which they use afterwards to encrypt the information before it is transmitted. In this case, the secrecy and the reliability of the transmission are managed separately in the two phases. The secrecy of the transmitted messages mainly depends on reliable information or reasonable and justifiable assumptions about the channel to the potential eavesdropper. Perfect state information about the channel to a passive eavesdropper is not a rational assumption. Thus, we introduce a deterministic model for the uncertainty about this channel, which yields a set of possible eavesdropper channels. We consider the optimization of worst-case rates in systems with multi-antenna Gaussian channels for both approaches. We study which transmit strategy can yield a maximum rate if we assume that the eavesdropper can always observe the corresponding worst-case channel that reduces the achievable rate for the secret transmission to a minimum. For both approaches, we show that the resulting max-min problem over the matrices that describe the multi-antenna system can be reduced to an equivalent problem over the eigenvalues of these matrices. We characterize the optimal resource allocation under a sum power constraint over all antennas and derive waterfilling solutions for the corresponding worst-case channel to the eavesdropper for a constraint on the sum of all channel gains. We show that all rates converge to finite limits for high signal-to-noise ratios (SNR), if we do not restrict the number of antennas for the eavesdropper. These limits are characterized by the quotients of the eigenvalues resulting from the Gramian matrices of both channels. For the low-SNR regime, we observe a rate increase that depends only on the differences of these eigenvalues for the direct-transmission approach. For the key generation approach, there exists no dependence from the eavesdropper channel in this regime. The comparison of both approaches shows that the superiority of an approach over the other mainly depends on the SNR and the quality of the eavesdropper channel. The direct-transmission approach is advantageous for low SNR and comparably bad eavesdropper channels, whereas the key generation approach benefits more from high SNR and comparably good eavesdropper channels. All results are discussed in combination with numerous illustrations.
Exploiting the randomness embedded in wireless channels, the physical-layer security techniques can provide additional security robustness against malicious eavesdropping at the physical layer of communication systems. To further this line of study, this thesis focuses on developing advanced multi-antenna transmission strategies for enhancing the secrecy performance of wireless communication systems. Assuming perfect channel feedback from the desired receiver and none from the eavesdropper, we consider beamforming with smartly injected artificial noise, aiming to confuse the eavesdropper whilst leaving the desired receiver unaffected. The system parameters are either chosen to be fixed for all transmissions, or adaptively adjusted based on the channel feedback from the desired receiver. For both cases, we provide novel and explicit design solutions for achieving the maximum throughput subject to a secrecy outage constraint. We then derive accurate approximations for the maximum throughput, and give new insights into the additional power cost for achieving a higher security level and the throughput gain of adaptive transmission over non-adaptive transmission. Having investigated the perfect feedback scenario, we then relax the feedback assumption and consider practical effects imposed by limited feedback constraints on the channel from the desired receiver to the transmitter. To deal with quantization error due to limited feedback whilst also accounting for unknown channel of the eavesdropper, we specify a connection outage constraint to guarantee the reliability of desired communication and a secrecy outage constraint to guard against malicious eavesdropping. Under dual outage constraints, we derive a novel optimized rate-adaptive transmission design to maximize the average secrecy throughput. New insights regarding the optimal power/feedback allocation and the number of feedback bits required for efficient quantization are obtained through our analysis. Our preceding study focused on isolated point-to-point transmission without any interference. However, the performance of contemporary wireless networks is often limited by interference from concurrent transmissions. To characterize the impact of interference on secure communication, we consider an interference-limited large-scale network. To confuse the malicious eavesdroppers, in addition to information-bearing signals, the transmitters are designed to generate artificial noise through either beamforming or antenna sectoring in a distributed manner. For both cases, using tools from stochastic geometry, we provide a statistical characterization of the reliability performance of desired communication and the secrecy performance against eavesdropping. We then investigate the networkwide secrecy throughput performance and optimize the transmit power allocation. Our analysis reveals interesting differences between the beamforming and sectoring approaches in terms of the optimal power allocation and the maximum networkwide secrecy throughput.
In this thesis, we study the physical layer security in downlink multi-user wireless networks. Traditionally, security has been addressed by cryptography at the higher layers of the communication stack. Security at the physical layer has been a major research topic in recent years. We study two different precoder designs alongside artificial noise (AN) to mitigate multi-user interference and deteriorate reception at the eavesdropper (Eve). We study the large scale analysis to calculate the secrecy sum-rate for these two cases and analyze the effect of AN on the system. First, we consider the worst case scenario, when eavesdropper's (Eve's) rate is not deteriorated by the interference caused by the legitimate users. Later, we investigate how interference from legitimate users would affect the large scale security sum rate. At the end, we assume more practical situation where the channel state information at the transmitter is not perfect due to feedback limitation and estimation error. We first study block diagonalization (BD) precoding. We derive the asymptotic secrecy sum-rate of BD for physical layer security in downlink multi-user wireless networks using artificial noise (AN). We first derive the secrecy sum-rate assuming that the channel of the eavesdropper is not known at the transmitter and derive the asymptotic secrecy sum-rate. Then we derive the optimum power allocation for asymptotic secrecy sum-rate. We design the system based on large scale analysis for different special cases. Simulation results show that the proposed precoders outperform channel inversion (CI) and regularized channel inversion (RCI) schemes at all SNRs. We also show that as the number of receive antennas of each user increases, the sum-rate gap between the proposed scheme and existing methods increases. We, then, study two methods of regularized block diagonalization (RBD) precoding using artificial noise (AN) for physical layer security in downlink multi-user MIMO wireless networks. We compare the complexity analysis of these two methods in terms of floating point operations per second (FLOPS). We derive the secrecy and asymptotic secrecy sum-rates for these schemes. We also derive closed form expressions for the asymptotic optimal power allocation to the AN signal and legitimate users. Simulations show RBD precoding outperform regularized channel inversion (RCI) in terms of secrecy sum-rate with a margin of 0.5 bits/s/Hz. Also, simulation results show our closed-form asymptotic secrecy sum-rates matches closely with the simulations for perfect and imperfect CSIT. We also study the effect of interference on BD precoding scheme. In the first two studies the worst case scenarios was considered and the effect of interference was not considered at the eavesdropper. We first derive the closed-form secrecy and asymptotic secrecy sum-rates for this scheme. Then, we derive the optimum power allocation. Using the derived optimum power allocation, we determine the conditions under which adding AN helps. We also investigate the optimum number of users, number of antennas at the eavesdropper, and the number of transmit antennas to fully study the conditions where using AN improves performance. We show that as the number of users increases or the number of receive antennas at the eavesdropper decreases, improved performance is achieved by allocating a smaller fraction of the total power to transmit the AN. The closed form expressions for the secrecy sum rate (SSR) and optimum power allocation closely match our simulation results. In Chapter 5, we study the effect of imperfect channel state information at the transmitter (CSIT) for each case studied in earlier chapters. We first study the effect of imperfect CSIT for BD precoding in both scenarios when interference does and does not exist at eavesdropper. We show that the imperfect CSI play the same role as interference for the desired users. Then, we study the effect of imperfect CSIT for RBD precoding when interference does not exist at eavesdropper. We show as channel estimation error increases, the secrecy sum-rate decreases. This is due to the fact that channel estimation error just affects the desired users' rate and has no effect on eavesdropper's performance. Finally, in the last chapter, we conclude our thesis and provide some road map for future--Author's abstract
Physical layer security (PLS) is a new paradigm aiming at securing communications between legitimate parties at the physical layer. Conventionally, achieving confidentiality in communication networks relies on cryptographic techniques such as public-key cryptography, secret-key distribution and symmetric encryption. Such techniques are deemed secure based on the assumption of limited computational abilities of a wiretapper. Given the relentless progress in computational capacities and the dynamic topology and proliferation of modern wireless networks, the relevance of the previous techniques in securing communications is more and more questionable and less and less reliable. In contrast to this paradigm, PLS does not assume a specific computational power at any eavesdropper, its premise to guarantee provable security via employing channel coding techniques at the physical layer exploiting the inherent randomness in most communication systems. In this dissertation, we investigate a particular aspect of PLS, which is secret-key agreement, also known as secret-sharing. In this setup, two legitimate parties try to distill a secret-key via the observation of correlated signals through a noisy wireless channel, in the presence of an eavesdropper who must be kept ignorant of the secret-key. Additionally, a noiseless public channel is made available to the legitimate parties to exchange public messages that are also accessible to the eavesdropper. Recall that key agreement is an important aspect toward realizing secure communications in the sense that the key can be used in a one-time pad scheme to send the confidential message. In the first part, our focus is on secret-sharing over Rayleigh fading quasi-static channels. We study the fundamental relationship relating the probability of error and a given target secret-key rate in the high power regime. This is characterized through the diversity multiplexing tradeoff (DMT) concept, that we define for our model and then characterize it. We show that the impact of the secrecy constraint is to reduce the effective number of transmit antennas by the number of antennas at the eavesdropper. Toward this characterization, we provide several schemes achieving the DMT and we highlight disparities between coding for the wiretap channel and coding for secret-key agreement. In the second part of the present work, we consider a fast-fading setting in which the wireless channels change during each channel use. We consider a correlated environment where transmit, legitimate receiver and eavesdropper antennas are correlated. We characterize the optimal strategy achieving the highest secret-key rate. We also identify the impact of correlation matrices and illustrate our analysis with various numerical results. Finally, we study the system from an energy-efficiency point of view and evaluate relevant metrics as the minimum energy required for sharing a secret-key bit and the wideband slope.
In this dissertation, we present novel physical (PHY) and cross-layer design guidelines and resource adaptation algorithms to improve the security and user experience in the future wireless networks. Physical and cross-layer wireless security measures can provide stronger overall security with high efficiency and can also provide better flexibility in response to the typically time-varying wireless environment. To better utilize limited wireless resources, we present approaches and techniques based on new performance criteria for wireless system optimization by thoroughly exploiting user information including its Quality of Service (QoS) requests and channel state information (CSI) to improve network throughput and security.
In recent years, there has been a growing research interest in wireless system security from PHY layer perspective. In a wiretap channel model introduced in the seminal work by Wyner, a sender “Alice” wishes to transmit a secret message to the intended receiver “Bob” in presence of a passive eavesdropper “Eve”. Existing works have characterized maximum achievable secrecy rate or secrecy capacity for single- and multiple-antenna systems by applying Gaussian signaling and secrecy code. Despite the impracticality of Gaussian input, its compact closed-form expression of mutual information motivated the wide use of Gaussian input assumption in theoretical analysis. In contrast to the Gaussian codebook, practical wiretap codes must consist of finite-alphabet symbols. Because of this constraint, the achievable secrecy rate for a finite-alphabet input scenario would differ from the secrecy rate achievable by a Gaussian codebook. In this thesis, we quantify the effect of finite discrete-constellation on instantaneous and ergodic secrecy rate of multiple-antenna wire-tap channels. Our results demonstrate substantial performance difference between systems involving finite-alphabet inputs and systems with Gaussian inputs. In addition, we investigate the secrecy performance of multi-terminal wiretap systems when a codebook based transmission beamforming is implemented. We characterize the secrecy outage probability of a communication link under eavesdropping. We consider a limited feedback scenario where the transmitter uses a pre-defined codebook known to both the transmitter and the receiver for beamforming, and analyze the secrecy outage probability of the link when it being eavesdropped.
Next, we consider the security of wireless multi-hop networks. Certainly, network security is an overarching issue that transcends all layers of communication protocol stack. Therefore, we tackle the problem of network security by proper user scheduling, routing, and resource management. Our approach of secrecy is complementary to the traditional cryptography based techniques. Together, they can provide stronger overall security and flexibility in selecting the desired solution and can reduce bandwidth and energy overhead as much as possible. Multi-hop wireless networks frequently employ the traditional routing and resource allocation strategies. Such strategies often fail to take into account the additional need for protection against security vulnerabilities due to the wireless medium. Thus, to address security concerns, we investigate the problem of eavesdropping control by proposing security-aware power allocation strategies and routing algorithms for multi-hop networks.
The future wireless infrastructure will consists of a plethora of heterogeneous collections of applications and services. A differentiated service structure would therefore be an essential part of the future wireless infrastructure. In addition to the security consideration of wireless network, we focus on the heterogeneous structure of the network. We study the problem of cross-layer resource allocation and admission control in a multiuser heterogeneous OFDMA network providing both QoS-constrained user servicesand best-effort services. Such heterogeneous networks provide service to both High-Priority (HP) users and Best-Effort (BE) users. By clustering subcarriers, efficient algorithms for cluster and power allocation have been proposed. Our strategy maximizes the total network utility of the BE users while satisfying the QoS request for maximum number of HP users. The feasibility of the resource allocation problem depends on the number of HP users in the network. Since a large number of demanding HP users would render the resource allocation problem infeasible, a joint admission control and resource allocation scheme could be an efficient way of tackling both problems with less overhead. By incorporating admission control, we propose an efficient and optimal joint admission control and resource allocation algorithm.
As modern infrastructure systems become increasingly more complex, we are faced with many new challenges in the area of information security. In this thesis we examine some approaches to security based on ideas from information theory. The protocols considered in this thesis, build upon the channel, a model for physical layer security proposed by A. Wyner in 1975. At a higher level, the protocols considered here can strengthen existing mechanisms for security by providing a new location based approach at the physical layer.
In the first part of this thesis, we extend the wiretap channel model to the case when there are multiple receivers, each experiencing a time varying fading channel. Both the scenario when each legitimate receiver wants a common message as well as the scenario when they all want separate messages are studied and capacity results are established in several special cases. When each receiver wants a separate independent message, an opportunistic scheme that transmits to the strongest user at each time, and uses Gaussian codebooks is shown to achieve the sum secrecy capacity in the limit of many users. When each receiver wants a common message, a lower bound to the capacity is provided, independent of the number of receivers.
In the second part of the thesis the role of multiple antennas for secure communication studied. We establish the secrecy capacity of the multi antenna wiretap channel (MIMOME channel), when the channel matrices of the legitimate receiver and eavesdropper are fixed and known to all the terminals. To establish the capacity, a new computable upper bound on the secrecy capacity of the wiretap channel is developed, which may be of independent interest. It is shown that Gaussian codebooks suffice to attain the capacity for this problem. For the case when the legitimate receiver has a single antenna (MISOME channel) a rank one transmission scheme is shown to attain the capacity. In the high signal-to-noise ratio (SNR) regime, it is shown that a capacity achieving scheme involves simultaneous diagonalization of the channel matrices using the generalized singular value decomposition and independently coding accross the resulting parallel channels. Furthermore a semi-blind masked beamforming scheme is studied, which transmits signal of interest in the subspace of the legitimate receiver's channel and synthetic noise in the orthogonal subspace. It is shown that this scheme is nearly optimal in the high SNR regime for the MISOME case and the performance penalty for the MIMOME channel is evaluated in terms of the generalized singular values. The behavior of the secrecy capacity in the limit of many antennas is also studied. When the channel matrices have i.i.d. C N(0, 1) entries, we show that (1) the secrecy capacity for the MISOME channel converges (almost surely) to zero if and only if the eavesdropper increases its antennas at a rate twice as fast as the sender (2) when a total of T G 1 antennas have to be allocated betweeen the sender and the receiver, the optimal allocation, which maximizes the number of eavesdropping antennas for zero secrecy capacity is 2 : 1.
In the final part of the thesis, we consider a variation of the wiretap channel where the sender and legitimate receiver also have access to correlated source sequences. They use both the sources and the structure of the underlying channel to extract secret keys. We provide general upper and lower bounds on the secret key rate and establish the capacity for the reversely degraded case. (Copies available exclusively from MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)
The use of wireless networks has grown significantly in contemporary \ntimes, and continues to develop further. The broadcast nature of \nwireless communications, however, makes them particularly vulnerable \nto eavesdropping. Unlike traditional solutions, which usually handle \nsecurity at the application layer, the primary concern of this \ndissertation is to analyze and develop solutions based on coding \ntechniques at the physical-layer. \n \nFirst, in chapter 2, we consider a scenario where a source node \nwishes to broadcast two confidential messages to two receivers, \nwhile a wire-tapper also receives the transmitted signal. This model \nis motivated by wireless communications, where individual secure \nmessages are broadcast over open media and can be received by any \nillegitimate receiver. The secrecy level is measured by the \nequivocation rate at the eavesdropper. We first study the general \n(non-degraded) broadcast channel with an eavesdropper, and present \nan inner bound on the secrecy capacity region for this model. This \ninner bound is based on a combination of random binning, and the \nGelfand-Pinsker binning. We further study the situation in which the \nchannels are degraded. For the degraded broadcast channel with an \neavesdropper, we present the secrecy capacity region. Our achievable \ncoding scheme is based on Cover's superposition scheme and random \nbinning. We refer to this scheme as the Secret Superposition Scheme. \nOur converse proof is based on a combination of the converse proof \nof the conventional degraded broadcast channel and Csiszar Lemma. We \nthen assume that the channels are Additive White Gaussian Noise and \nshow that the Secret Superposition Scheme with Gaussian codebook is \noptimal. The converse proof is based on Costa's entropy power \ninequality. Finally, we use a broadcast strategy for the slowly \nfading wire-tap channel when only the eavesdropper's channel is \nfixed and known at the transmitter. We derive the optimum power \nallocation for the coding layers, which maximizes the total average \nrate. \n \n \nSecond, in chapter 3 , we consider the \nMultiple-Input-Multiple-Output (MIMO) scenario of a broadcast \nchannel where a wiretapper also receives the transmitted signal via \nanother MIMO channel. First, we assume that the channels are \ndegraded and the wiretapper has the worst channel. We establish the \ncapacity region of this scenario. Our achievability scheme is the \nSecret Superposition Coding. For the outerbound, we use notion of \nthe enhanced channels to show that the secret superposition of \nGaussian codes is optimal. We show that we only need to enhance the \nchannels of the legitimate receivers, and the channel of the \neavesdropper remains unchanged. We then extend the result of the \ndegraded case to a non-degraded case. We show that the secret \nsuperposition of Gaussian codes, along with successive decoding, \ncannot work when the channels are not degraded. We develop a Secret \nDirty Paper Coding scheme and show that it is optimal for this \nchannel. We then present a corollary generalizing the capacity \nregion of the two receivers case to the case of multiple receivers. \nFinally, we investigate a scenario which frequently occurs in the \npractice of wireless networks. In this scenario, the transmitter and \nthe eavesdropper have multiple antennae, while both intended \nreceivers have a single antenna (representing resource limited \nmobile units). We characterize the secrecy capacity region in terms \nof generalized eigenvalues of the receivers' channels and the \neavesdropper's channel. We refer to this configuration as the MISOME \ncase. We then present a corollary generalizing the results of the \ntwo receivers case to multiple receivers. In the high SNR regime, we \nshow that the capacity region is a convex closure of rectangular \nregions. \n \n \nFinally, in chapter 4, we consider a K-user secure Gaussian \nMultiple-Access-Channel with an external eavesdropper. We establish \nan achievable rate region for the secure discrete memoryless MAC. \nThereafter, we prove the secrecy sum capacity of the degraded \nGaussian MIMO MAC using Gaussian codebooks. For the non-degraded \nGaussian MIMO MAC, we propose an algorithm inspired by the \ninterference alignment technique to achieve the largest possible \ntotal Secure-Degrees-of-Freedom . When all the terminals are \nequipped with a single antenna, Gaussian codebooks have shown to be \ninefficient in providing a positive S-DoF. Instead, we propose a \nnovel secure coding scheme to achieve a positive S-DoF in the single \nantenna MAC. This scheme converts the single-antenna system into a \nmultiple-dimension system with fractional dimensions. The \nachievability scheme is based on the alignment of signals into a \nsmall sub-space at the eavesdropper, and the simultaneous separation \nof the signals at the intended receiver. We use tools from the field \nof Diophantine Approximation in number theory to analyze the \nprobability of error in the coding scheme. We prove that the total \nS-DoF of fracK−1K can be achieved for almost all channel \ngains. For the other channel gains, we propose a multi-layer coding \nscheme to achieve a positive S-DoF. As a function of channel gains, \ntherefore, the achievable S-DoF is discontinued.