MIMO-based Friendly Jamming and Interference Management Techniques for Secure Wireless Communications

Abstract

The ever-increasing growth of wireless systems has made them an essential part of our daily life. People rely heavily on wireless networks for communications and to conduct critical transactions from their mobile devices, including financial transactions, access to health records, etc. The proliferation of wireless communication devices opens the door for many security breaches, ranging from eavesdropping to jamming attacks. Such a disadvantage stems from the broadcast nature of wireless transmissions, which creates an exposed environment. In this dissertation, we focus on eavesdropping attacks. While cryptographic techniques can be used to thwart eavesdropping attacks and enable secure wireless communications, they are not sufficient to protect the lower-layer headers of a packet (i.e., PHY and MAC headers). Hence, even though the secret message is encrypted, these unencrypted headers can be exploited by an adversary to extract invaluable information and initiate malicious attacks (e.g., traffic classification). Physical-layer (PHY-layer) security has been introduced as a promising candidate to prevent attacks that exploit unencrypted lower layer headers. PHY-layer security techniques typically rely on injecting an intentional interference into the medium so as to confuse nearby eavesdroppers (Eve). Specifically, a legitimate transmit-receive (Alice-Bob) pair generates a bogus signal, namely friendly jamming (FJ), along with the information signal, to increase interference at Eve(s) but without affecting the legitimate receiver (Bob). Depending on which end of a legitimate link is responsible for generating the FJ signal, two types of FJ techniques exist: transmitter-based (TxFJ) and receiver-based (RxFJ). In this dissertation, we propose to advance the state-of-art in PHY-layer security by considering multi-link scenarios, including multi-user multiple-input multiple-output (MU-MIMO) and peer-to-peer (P2P) networks. Specifically, we consider a scenario where one or more external Eve(s) attempt to snoop on communications of various links. In such networks, transmission of one link may be interfered with neighboring links' transmissions. Thus, special care must be dedicated to handling interference. In our first contribution in this dissertation, we consider a P2P network tapped by external Eve(s) in which each Alice-Bob pair conceals its communications using TxFJ. TxFJ is realized at Alice side using MIMO precoding. The goal is to design the precoders for both information and TxFJ signals at all Alices so as to maximize a given utility (e.g., sum of communication rates) while preventing eavesdropping elsewhere. Because legitimate links do not cooperate with each other and there is no centralized authority to perform optimization, every link selfishly aims at maximizing its secrecy rate. Using non-cooperative game theory, we design a distributed method for maximizing the sum of secrecy rates. Under the exact knowledge of eavesdropping channels, we show that our distributed method has a comparable secrecy sum-rate to a centralized approach. In our next contribution, we focus on employing practical precoders in our design for a P2P network. Specifically, we employed a zero-forcing-based (ZF-based) precoder for the TxFJ of each Alice-Bob pair in a P2P network. We also assume that each link has a certain rate demand to be satisfied. In such a scenario, even though the non-cooperative game designed for this P2P network is shown to be convergent to its unique Nash Equilibrium (NE), there is still no guarantee that the resulting NE is Pareto-optimal. Hence, we propose a modified price-based game, in which each link is penalized for generating interference on other legitimate links. We show that the price-based game converges to the Pareto-optimal point of secrecy rate region. We then leverage mixed-strategy games to provide solutions that are robust to uncertainties in knowledge of eavesdropping channels. The proposed ZF-based design of precoders is also implemented on software-defined radios to assess its performance on a single link in real-world scenarios. In another contribution of this dissertation, we consider to further enhance the secrecy of each link in a P2P network by equipping each receiver with RxFJ. Hence, in addition to the power allocation between TxFJ and information signals, we optimize RxFJ power as well. We show that by using RxFJ at each Bob, we could leverage the well-established concept of concave games, which compared to non-convex games enjoy more simplified game-theoretic analysis. We derive sufficient conditions under which the game admits a unique NE. We also propose another version of our power control algorithm that can be implemented asynchronously, making it robust to transmission delays in the network. In our last contribution, we consider the downlink of a MU-MIMO network in the presence of an external Eve. No knowledge of Eve's location is assumed at the access point. The network is studied in underloaded and overloaded conditions. In an underloaded (overloaded) network, the number of antennas at the access point is larger (smaller) than the total number of downlink users' antennas. In the overloaded setting, traditional methods of creating TxFJ, such as ZF-based methods, are infeasible. We propose a linear precoding scheme that relaxes such infeasibility in overloaded MU-MIMO networks. In the worst-case scenario where Eve has knowledge of the channels between access point and downlink users, we show that our method imposes the most stringent condition on the number of antennas required at Eve to cancel out TxFJ signals. We also show that choosing the number of independent streams to be sent to downlink users has an important role in achieving a tradeoff between security, reliability, and the achievable rate.