Information Theoretic Secrecy for Some Multiuser Wireless Communication Channels

The widespread use of multiuser wireless communications has been raising the demand for higher data rates and until now that has been the main design focus in wireless systems. This design focus, however, neglects the critical issue of communications security inherent in wireless systems that present a channel that is very easy for people of criminal intent to tap. Existing approaches leave this important issue to the higher layers of the protocol hierarchies, yet the need to deal with it in lower layers is imminent as the security of many cryptographic algorithms is hard to evaluate and has caused disappointment in the past. In addition, there is rising interest in large networks of low-complexity transmitters such as sensor nodes and RF-ID tags that do not have room for complicated and computationally intensive cryptographic algorithms. Historically, communication secrecy has been evaluated using one of two approaches: information theory or computational complexity. Throughout this thesis, we look at the information theoretic measures of secrecy to find provable secure communication schemes, and try to bring a joint approach to improve both security and throughput. We start by examining the Gaussian Multiple-Access Wire-Tap Channel (GMAC-WT), and an eavesdropper who has access to a degraded version of the legitimate receiver's signal. For this scenario, we first define two suitable secrecy measures, called "collective" and "individual" secrecy measures. We first look at coding schemes to achieve a given amount of secrecy in the presence of this eavesdropper. We find achievable secrecy rate regions for each constraint, and show that we can achieve the secrecy sum-capacity for this scenario. We then look at a more general scenario where the eavesdropper does not necessarily have a degraded version of the receiver's signal. In this more general case, we consider the "collective" constraints, as this constraint allows the system to achieve a larger rate region by placing trust on all users; and the "perfect secrecy" case, since this is in general a more appropriate constraint for practical purposes. We find an achievable secrecy rate region, and the power allocations that maximize the achievable perfect secrecy sum-rate. We also introduce a novel scheme, which we call "cooperative jamming," where users whose signals are vulnerable to decoding by the eavesdropper jam the eavesdropper, achieving higher secrecy rates for the remaining users. We also find an outer bound on the achievable secrecy sum-rate, which is shown to coincide with the achievable secrecy sum-rate for the degraded scenario. Next, we consider the Gaussian Two-Way Wire-Tap Channel (GTW-WT) and the Binary Additive Two-Way Wire-Tap Channel (BATW-WT). For these channels, we find achievable secrecy regions, and show that utilizing the same channel, in addition to the transmitters knowing their own self-interference, allows for higher secrecy sum-rates than the GMAC-WT. We also investigate cooperative jamming and find the optimum transmit/jam powers for the two users. Finally, we examine the ergodic block-fading case for the two user GMAC-WT. We find the secrecy sum-rate maximizing power allocations under a long-term power constraint, and see that unlike the standard GMAC where it is optimal for only one user to be transmitting at any given time, in some cases it is better for both users to be transmitting. We also consider cooperative jamming for this scenario, and give numerical results showing the improvement achievable for the secrecy sum-rate. This thesis is among the first to point out the advantages that the wireless medium brings to securing information transfer...