Robust Physical Layer Security with Channel Knowledge Uncertainty

Abstract

Next generation wireless systems such as 5G and beyond are key to meeting the increasing demands of both the cellular and other ubiquitously connected, but highly mobile devices, in internet of things (IoT) networks. However, this need for instant interconnectivity for (perhaps moving) neighboring devices or machines may not tolerate latency in channel estimation and security methods (e.g., cryptography) which require secret-key-sharing infrastructures and authentication. For the stated reasons and more, there has been a rising interest in physical layer security (PHY-SEC) as an alternative approach. PHY-SEC relies on leveraging the inherent randomness of the wireless medium such as fading or noise. Moreover, for networks that allow so, PHY-SEC is tipped to serve as an extra layer of security beyond the key-sharing methods that are currently used at the upper layers of the Open Systems Interconnection (OSI) model. The goal of this dissertation is to devise transmission algorithms and characterize fundamental limits of information theoretic PHY-SEC. In particular, the aim is to maximize the transmission data rates between the legitimate transceivers, while not leaking any information to eavesdroppers. The majority of research developments on PHY-SEC have been made under the assumptions that the wireless channel state information (CSI) is available at the transmitters (CSIT). This, however, is not practically feasible because any eavesdroppers in the vicinity cannot be expected to cooperate with the legitimate transmitters in estimating CSI and feeding it back to the transmitters. It thus remains imperative to design transmission schemes that can achieve high secure data rates with little to no CSIT. Moreover, most of the classic research results model the wireless medium as a discrete memoryless channel or DMC; however, this disregards the inherent presence of intersymbol interference (ISI) due to multipath propagation. Furthermore, the majority of existing literature and texts assume that the multi-user interference network topology (e.g., $K$-user network) is fully connected, i.e., that every receiver observes its intended signal and all the other ($K-1$) signals as interference. In reality, however, the topology may be partially connected (i.e., such that each receiver is only connected to an arbitrary subset of transmitters and vice versa) due to the inherent wireless channel impediments such as pathloss and block fading. To this end, in this thesis, we make the following research contributions: a) We devise transmission schemes to exploit the interplay between the relative number of antennas at each terminal (i.e., multiple-input multiple-output or MIMO channels) and the (often non-desirable) ISI due to multi-path fading for the point to point channel in presence of a passive eavesdropper. We expand the concept of ISI exploitation for PHY-SEC to multi-user interference networks. In particular, we propose a transmission scheme that leverages tractable structures of the received signals under the orthogonal frequency division multiplexing (OFDM) transmission for fully connected multi-user interference channels. b) We explore the interplay between network topologies (graph theory) and the inherent impediments of the wireless channel such as block fading and pathloss. In particular, we study the problem of partially connected multi-user interference networks. We refer to this framework where transmitters do not have any channel knowledge except the network topology as topological interference management with confidential messages (TIM-CM). c) We study the trade-off between achieving high data rates versus allowing some leakage of less privacy-sensitive attributes of data to the eavesdroppers. More specifically, we introduce and study a problem referred to as the latent variable wiretap channel (LV-WTC). Several open problems and future research directions that originate from the problems formulated herein are also discussed.