<p><em>Imbulgoda Liyangahawatte, Gihan Janith Mendis Ph.D., Purdue University, May</em></p> <p><em>2023. Deep learning for securing critical infrastructure with the emphasis on power</em></p> <p><em>systems and wireless communication. Major Professor: Dr. Jin Kocsis.</em></p> <p><br></p> <p><em>Critical infrastructures, such as power systems and communication</em></p> <p><em>infrastructures, are of paramount importance to the welfare and prosperity of</em></p> <p><em>modern societies. Therefore, critical infrastructures have a high vulnerability to</em></p> <p><em>attacks from adverse parties. Subsequent to the advancement of cyber technologies,</em></p> <p><em>such as information technology, embedded systems, high-speed connectivity, and</em></p> <p><em>real-time data processing, the physical processes of critical infrastructures are often</em></p> <p><em>monitored and controlled through cyber systems. Therefore, modern critical</em></p> <p><em>infrastructures are often viewed as cyber-physical systems (CPSs). Incorporating</em></p> <p><em>cyber elements into physical processes increases efficiency and control. However, it</em></p> <p><em>also increases the vulnerability of the systems to potential cybersecurity threats. In</em></p> <p><em>addition to cyber-level attacks, attacks on the cyber-physical interface, such as the</em></p> <p><em>corruption of sensing data to manipulate physical operations, can exploit</em></p> <p><em>vulnerabilities in CPSs. Research on data-driven security methods for such attacks,</em></p> <p><em>focusing on applications related to electrical power and wireless communication</em></p> <p><em>critical infrastructure CPSs, are presented in this dissertation. As security methods</em></p> <p><em>for electrical power systems, deep learning approaches were proposed to detect</em></p> <p><em>adversarial sensor signals targeting smart grids and more electric aircraft.</em></p> <p><em>Considering the security of wireless communication systems, deep learning solutions</em></p> <p><em>were proposed as an intelligent spectrum sensing approach and as a primary user</em></p> <p><em>emulation (PUE) attacks detection method on the wideband spectrum. The recent</em></p> <p><em>abundance of micro-UASs can enable the use of weaponized micro-UASs to conduct</em></p> <p><em>physical attacks on critical infrastructures. As a solution for this, the radio</em></p> <p><em>frequency (RF) signal-analyzing deep learning method developed for spectrum</em></p> <p><em>sensing was adopted to realize an intelligent radar system for micro-UAS detection.</em></p> <p><em>This intelligent radar can be used to provide protection against micro-UAS-based</em></p> <p><em>physical attacks on critical infrastructures.</em></p>

<p dir="ltr"> The general design goal of low latency communication systems is to minimize the end-to-end delay while attaining the predefined reliability and throughput requirements. The burgeoning demand for low latency communications motivates a renewed research interest of the tradeoff between delay, throughput, and reliability. In this dissertation research, we consider slotted-based systems and explore the potential advantages of the so-called sequential codes in low latency network communications.</p><p dir="ltr"> The first part of this dissertation analyzes the exact error probability of random linear streaming codes (RLSCs) in the large field size regime over the stochastic independently and identically distributed (i.i.d.) symbol erasure channels (SECs). A closed-form expression of the error probability <i>p</i>_<em>e</em> of large-field-size RLSCs is derived under, simultaneously, the finite memory length α and decoding deadline Δ constraints. The result is then used to examine the intricate tradeoff between memory length (complexity), decoding deadline (delay), code rate (throughput), and error probability (reliability). Numerical evaluation shows that under the same code rate and error probability requirements, the end-to-end delay of RLSCs is 40-48% of that of the optimal block codes (i.e., MDS codes). This implies that switching from block codes to streaming codes not only eliminates the queueing delay completely (which accounts for the initial 50% of the delay reduction) but also improves the reliability (which accounts for the additional 2-10% delay reduction).</p><p dir="ltr"> The second part of this dissertation focuses on the asymptotics of the error probability of RLSCs in the same system model of the first part. Two important scenarios are analyzed: (i) tradeoff between Δ and <i>p</i>_<em>e</em> under infinite α; and (ii) tradeoff between α and <i>p</i>_<em>e</em> under infinite Δ. In the first scenario, the asymptote of <i>p</i>_<em>e</em>(Δ) is shown to be <i>ρ</i>Δ^-1.5<i>e</i>^-^<em>η</em>^Δ. The asymptotic power term Δ^-1.5 of RLSCs is a strict improvement over the Δ^-0.5 term of random linear block codes. A pair of upper and lower bound on the asymptotic constant <i>ρ</i> is also derived, which are tight (i.e., identical) for one specific class of SECs. In the second scenario, a refine approximation is proposed by computing the parameters in a multiterm asymptotic form, which closely matches the exact error probability even for small memory length (≈ 20). The results of the asymptotics can be further exploited to find the <i>c</i>-optimal memory length <i>α</i>_<em>c</em>^*(Δ), which is defined as the minimal memory length α needed for the resulting <i>p</i>_<em>e</em> to be within a factor of <i>c</i>>1 of the best possible <i>p</i>_<em>e</em>^<em>*</em>_{<em> </em>}for any Δ, an important piece of information for practical implementation.</p><p dir="ltr"> Finally, we characterize the channel dispersions of RLSCs and MDS block codes, respectively. New techniques are developed to quantify the channel dispersion of sequential (non-block-based) coding, the first in the literature. The channel dispersion expressions are then used to compare the levels of error protection between RLSCs and MDS block codes. The results show that if and only if the target error probability <i>p</i>_<em>e</em> is smaller than a threshold (≈ 0.1774), RLSCs offer strictly stronger error protection than MDS block codes, which is on top of the already significant 50% latency savings of RLSCs that eliminate the queueing delay completely.</p>

<p dir="ltr">With the increasing availability of inexpensive computing power in wireless radio network nodes, machine learning based models are being deployed in operations that traditionally relied on rule-based or statistical methods. Contemporary high bandwidth networks enable easy availability of significant amounts of training data in a comparatively short time, aiding in the development of better deep learning models. Specialized deep learning models developed for wireless networks have been shown to consistently outperform traditional methods in a variety of wireless network applications.</p><p><br></p><p dir="ltr">We aim to address some of the unique challenges inherent in the wireless radio communication domain. Firstly, as data is transmitted over the air, data privacy and adversarial attacks pose heightened risks. Secondly, due to the volume of data and the time-sensitive nature of the processing that is required, the speed of the machine learning model becomes a significant factor, often necessitating operation within a latency constraint. Thirdly, the impact of diverse and time-varying wireless environments means that any machine learning model also needs to be generalizable. The increasing computing power present in wireless nodes provides an opportunity to offload some of the deep learning to the edge, which also impacts data privacy.</p><p><br></p><p dir="ltr">Towards this goal, we work on deep learning methods that operate along different aspects of a wireless network—on network packets, error prediction, modulation classification, and channel estimation—and are able to operate within the latency constraint, while simultaneously providing better privacy and security. After proposing solutions that work in a traditional centralized learning environment, we explore edge learning paradigms where the learning happens in distributed nodes.</p>

The thesis describes a model of electromagnetic pulses from sources between 100MHz and 100GHz for use in the development of Knowledge-Based systems. Each pulse, including its modulations, is described as would be seen by a definable receiving system. The model has been validated against a range of Knowledge-Based systems including a neural network, a Learning systems and an Expert system.

<p>The swift emergence and wide-ranging utilization of machine learning (ML) across various industries, including healthcare, transportation, and robotics, have underscored the escalating need for efficient, scalable, and privacy-preserving solutions. Recognizing this, we present an integrated examination of three novel frameworks, each addressing different aspects of distributed learning and privacy issues: Two Timescale Hybrid Federated Learning (TT-HF), Delay-Aware Federated Learning (DFL), and Differential Privacy Hierarchical Federated Learning (DP-HFL). TT-HF introduces a semi-decentralized architecture that combines device-to-server and device-to-device (D2D) communications. Devices execute multiple stochastic gradient descent iterations on their datasets and sporadically synchronize model parameters via D2D communications. A unique adaptive control algorithm optimizes step size, D2D communication rounds, and global aggregation period to minimize network resource utilization and achieve a sublinear convergence rate. TT-HF outperforms conventional FL approaches in terms of model accuracy, energy consumption, and resilience against outages. DFL focuses on enhancing distributed ML training efficiency by accounting for communication delays between edge and cloud. It also uses multiple stochastic gradient descent iterations and periodically consolidates model parameters via edge servers. The adaptive control algorithm for DFL mitigates energy consumption and edge-to-cloud latency, resulting in faster global model convergence, reduced resource consumption, and robustness against delays. Lastly, DP-HFL is introduced to combat privacy vulnerabilities in FL. Merging the benefits of FL and Hierarchical Differential Privacy (HDP), DP-HFL significantly reduces the need for differential privacy noise while maintaining model performance, exhibiting an optimal privacy-performance trade-off. Theoretical analysis under both convex and nonconvex loss functions confirms DP-HFL’s effectiveness regarding convergence speed, privacy performance trade-off, and potential performance enhancement with appropriate network configuration. In sum, the study thoroughly explores TT-HF, DFL, and DP-HFL, and their unique solutions to distributed learning challenges such as efficiency, latency, and privacy concerns. These advanced FL frameworks have considerable potential to further enable effective, efficient, and secure distributed learning.</p>

<p>The last few decades have witnessed unprecedented growth in multiple areas of electronics spanning low-power sensing, intelligent computing and high-speed wireless connectivity. In the foreseeable future, there would be hundreds of billions of computing devices, sensors, things and people, wherein the technology will become intertwined with our lives through continuous interaction and collaboration between humans and machines. Such human-centric ideas give rise to the concept of internet of bodies (IoB), which calls for novel and energy-efficient techniques for sensing, processing and secure communication for resource-constrained IoB nodes.As we have painfully learnt during the pandemic, point-of-care diagnostics along with continuous sensing and long-term connectivity has become one of the major requirements in the healthcare industry, further emphasizing the need for energy-efficiency and security in the resource-constrained devices around us.</p> <p>  </p> <p>  With this vision in mind, I’ll divide this dissertation into the following chapters. The first part (Chapter 2) will cover time-domain sensing techniques which allow inherent energy-resolution scalability, and will show the fundamental limits of achievable resolution. Implementations will include 1) a radiation sensing system for occupational dosimetry in healthcare and mining applications, which can achieve 12-18 bit resolution with 0.01-1 µJ energy dissipation, and 2) an ADC-less neural signal acquisition system with direct Analog to Time Conversion at 13pJ/Sample. The second part (Chapters 3 and 4) of this dissertation will involve the fundamentals of developing secure energy-efficient electro-quasistatic (EQS) communication techniques for IoB wearables as well as implants, and will demonstrate  2 examples: 1) Adiabatic Switching for breaking the αCV^2f limit of power consumption in capacitive voltage mode human-body communication (HBC), and 2) Bi-Phasic Quasistatic Brain Communication (BP-QBC) for fully wireless data transfer from a sub-6mm^3, 2 µW brain implant. A custom modulation scheme, along with adiabatic communication enables wireline-like energy efficiencies (<5pJ/b) in HBC-based wireless systems, while the BP-QBC node, being fully electrical in nature, demonstrates sub-50pJ/b efficiencies by eliminating DC power consumption, and by avoiding the transduction losses observed in competing technologies, involving optical, ultrasound and magneto-electric modalities. Next in Chapter 5, we will show an implementation of a reconfigurable system that would include 1) a human-body communication transceiver and 2) a traditional wireless (MedRadio) transceiver on the same integrated circuit (IC), and would demonstrate methods to switch between the two modes by detecting the placement of the transmitter and receiver devices (on-body/away from the body). Finally, in Chapter 6, we shall show a technique of augmenting security in resource-constrained devices through authentication using the Analog/RF properties of the transmitter, which are usually discarded as non-idealities in a digital transceiver chain. This method does not require any additional hardware in the transmitter, making it an extremely promising technique to augment security in highly resource-constrained scenarios. Such energy-efficient intelligent sensing and secure communication techniques, when combined with intelligent in-sensor-analytics at the resource-constrained nodes, can potentially pave the way for perpetual, and even batteryless systems for next-generation IoT, IoB and healthcare applications.</p>