Defense Notices

**Currently under security review**

Metagenomic sequencing has become a central paradigm for studying complex microbial communities and their interactions with the host, with emerging applications in clinical prediction and disease modeling. In this work, we first investigate two representative application scenarios: predicting immune checkpoint inhibitor response in non-small cell lung cancer using gut microbial signatures, and characterizing host–microbiome interactions in neonatal systems. The proposed reference-free neural network captures both compositional and functional signals without reliance on reference genomes, while the neonatal study demonstrates how environmental and genetic factors reshape microbial communities and how probiotic intervention can mitigate pathogen-induced immune activation.

These studies highlight both the promise and the inherent difficulty of metagenomic analysis: transforming raw sequencing data into clinically actionable insights remains an algorithmically fragmented and computationally intensive process. This challenge arises from two key limitations: the lack of a unified algorithmic foundation for sequence alignment and the absence of systematic approaches for selecting and organizing analytical tools. Motivated by these challenges, we present a unified computational framework for metagenomic analysis that integrates complementary algorithmic and systems-level solutions.

First, to resolve fragmentation at the alignment level, we develop the Versatile Alignment Toolkit (VAT), a unified algorithmic system for biological sequence alignment across diverse applications. VAT introduces an asymmetric multi-view k-mer indexing scheme that integrates multiple seeding strategies within a single architecture and enables dynamic seed-length adjustment via longest common prefix (LCP)–based inference without re-indexing. A flexible seed-chaining mechanism further supports diverse alignment scenarios, including collinear, rearranged, and split alignments. Combined with a hardware-efficient in-register bitonic sorting algorithm and dynamic index-loading strategy, VAT achieves high efficiency and broad applicability across read mapping, homology search, and whole-genome alignment. Second, to address the challenge of tool selection and pipeline construction, we develop SNAIL, a natural language processing system for automated recognition of bioinformatics tools from large-scale and rapidly growing scientific literature. By integrating XGBoost and Transformer-based models such as SciBERT, SNAIL enables structured extraction of analytical tools and supports automated, reproducible pipeline construction.

Together, this work establishes a unified framework that is grounded in real-world applications and addresses key bottlenecks in metagenomic analysis, enabling more efficient, scalable, and clinically actionable workflows.

Conventional computer vision relies on spatially resolved, human-interpretable images, which inherently expose sensitive information and raise privacy concerns. In this study, we explore an alternative paradigm based on lensless imaging, where scenes are captured as diffraction patterns governed by the point spread function (PSF). Although unintelligible to humans, these measurements encode structured, distributed information that remains useful for computational inference. 

We propose a unified framework for privacy-preserving vision that operates directly on lensless sensor measurements by leveraging their frequency-domain and phase-encoded properties. The framework is developed along two complementary directions. First, we enable reconstruction-free inference by exploiting the intrinsic obfuscation of lensless data. We show that semantic tasks such as classification can be performed directly on diffraction patterns using models tailored to non-local, phase-scrambled representations. We further design lensless-aware architectures and integrate them into practical pipelines, including a Swin Transformer-based steganographic framework (DiffHide) for secure and imperceptible information embedding. To assess robustness, we formalize adversarial threat models and develop defenses against learning-based reconstruction attacks, particularly GAN-driven inversion. Second, we investigate the limits of privacy by studying the reconstructability of lensless measurements without explicit knowledge of the forward model. We develop learning-based reconstruction methods that approximate the inverse mapping and analyze conditions under which sensitive information can be recovered. Our results demonstrate that lensless measurements enable effective vision tasks without reconstruction, while providing a principled framework to evaluate and mitigate privacy risks. 

Coherent wavelength-division multiplexed (WDM) optical fiber systems have become the primary transmission technology for high-capacity data networks, driven by the explosive bandwidth demand of cloud computing, streaming services, and large-scale artificial intelligence training infrastructure. This dissertation investigates two fundamental aspects of digital coherent fiber optic systems under the unifying theme of source and monitoring: the design of multi-wavelength optical sources compatible with high-order coherent detection, and the leveraging of fiber Kerr-effect nonlinearity at the coherent receiver to perform physical-layer link health monitoring and to assess inherent security vulnerabilities — both achieved through digital signal processing of the received complex optical field without dedicated hardware.

We begin by addressing the multi-wavelength transmitter challenge in WDM coherent systems. Existing quantum-dot, quantum-dash, and quantum-well based optical frequency comb (OFC) sources share a common limitation: individual comb line linewidths in the tens of MHz range caused by low output power levels of 1–20 mW, making them incompatible with high-order coherent detection. We demonstrate coherent system application of a single-section InGaAsP QW Fabry-Perot laser diode with greater than 120 mW optical power at the fiber pigtail and 36.14 GHz mode spacing. The high optical power per mode produces Lorentzian equivalent linewidths below 100 kHz — compatible with 16-QAM carrier phase recovery without optical phase locking. Experimental results obtained using a commercial Ciena WaveLogic-Ai coherent transceiver demonstrate 20-channel WDM transmission over 78.3 km of standard single-mode fiber with all channels below the HD-FEC threshold of 3.8 × 10⁻³ at 30 GBaud differential-coded 16-QAM, corresponding to an aggregate capacity of 2.15 Tb/s from a single laser device.

After investigating the QW Fabry-Perot laser as a multi-wavelength source for coherent WDM transmission, we leverage the coherent receiver DSP to exploit fiber Kerr-effect nonlinearity for longitudinal power profile estimation, enabling reconstruction of the signal power distribution P(z) along the full multi-span link without dedicated hardware or traffic interruption. We propose a modified enhanced regular perturbation (ERP) method that corrects two independent physical error sources of the standard RP1 least-squares baseline: the accumulated nonlinear phase rotation, and the dispersion-mediated phase-to-intensity conversion — a second bias source not addressed by prior methods. The RP1 method produces mean absolute error (MAE) that scales quadratically with span count, growing to 1.656 dB at 10 spans and 3 dBm. The modified ERP reduces this to 0.608 dB — an improvement that grows consistently with link length, confirming increasing advantage in the long-haul regime. Extension to WDM through an XPM-aware per-channel formulation achieves MAE of 0.113–0.419 dB across 150–500 km link lengths.

In addition to its role in enabling DSP-based longitudinal power profile estimation, the fiber Kerr-effect nonlinearity is shown to give rise to an inherent physical-layer security vulnerability in coherent WDM systems. We show that an eavesdropper co-tenanting a shared fiber — transmitting a continuous-wave probe at a wavelength adjacent to the legitimate signal — can capture the XPM-induced waveform at the fiber output and apply a bidirectional gated recurrent unit neural network, trained on split-step Fourier method simulation data, to reconstruct the transmitted symbol sequence without physical fiber access and without perturbing the legitimate signal. This eavesdropping mechanism is experimentally validated using a commercial Ciena WaveLogic-Ai coherent transceiver for ASK, BPSK, QPSK, and 16-QAM modulation formats at 4.26 GBaud and 8.56 GBaud over one- and two-span 75 km fiber systems, achieving zero symbol errors under high-OSNR conditions. Noise-aware training over OSNR from 20 to 60 dB maintains symbol error rate below 10⁻² for OSNR above 25–30 dB.

Together, these three contributions demonstrate that the coherent fiber optic system is a versatile physical instrument extending well beyond its role as a data transmission medium. The coherent receiver infrastructure — deployed for high-order modulation and data recovery — simultaneously enables the high-power OFC laser to serve as a practical multi-wavelength transmitter source, and provides the complex field measurement capability through which fiber Kerr-effect nonlinearity can be exploited constructively for distributed link monitoring and, as a direct consequence, reveals an inherent physical-layer security exposure in shared fiber infrastructure. This unified perspective on the coherent system as both a transmission platform and a general-purpose measurement instrument has direct relevance to the design of spectrally efficient, self-monitoring, and physically secure optical interconnects for next-generation AI computing networks.

Advances in sensing, communication, and intelligent control have transformed power systems into data-driven smart grids, where forecasting and intelligent decision-making are essential components. Modern smart grids include distributed energy resources (DERs), renewable generation, battery energy storage systems, and large numbers of grid-edge devices that continuously generate time-series data. At the same time, increasing renewable penetration introduces substantial uncertainty in generation, net load, and market operations, while communication networks impose bandwidth, latency, and reliability constraints on timely data delivery. This dissertation addresses how time-series forecasting, data compression, and task-oriented wireless communication can be jointly designed for smart grid applications.

First, we study weather-aware distributed energy management in prosumer-centric microgrids and show that incorporating day-ahead weather information into decision-making improves battery dispatch and reduces the impact of renewable uncertainty. Second, we introduce forecasting-aware energy management in both wholesale and retail electricity markets, highlighting how renewable generation forecasting affects pricing, scheduling, and uncertainty mitigation. Third, we develop and evaluate deep learning methods for renewable generation forecasting, showing that Transformer-based models outperform recurrent baselines such as RNN and LSTM for wind and solar prediction tasks.

Building on this forecasting foundation, we develop a communication-efficient forecasting framework in which high-dimensional smart grid measurements are compressed into low-dimensional latent representations before transmission. This framework is extended into a task-oriented communication system that jointly optimizes data relevance and information timeliness, so that the receiver obtains compressed updates that remain useful for downstream forecasting tasks. Finally, we extend this framework to a distributed multi-node uplink setting, where multiple grid sensors share a bandwidth-limited channel, and develop scheduling policy that improves both the timeliness and task-relevance of received updates.

Alzheimer’s Disease (AD) is a progressive, irreversible neurodegenerative disorder and the leading cause of dementia worldwide, affecting an estimated 55 million people globally. The window of opportunity for intervention is demonstrably narrow, making reliable early-stage detection a clinical and scientific imperative. While current diagnostic techniques such as neuroimaging and cerebrospinal fluid (CSF) biomarkers carry well-defined limitations in scalability, cost, and access equity, speech has emerged as a compelling non-invasive proxy for cognitive function evaluation.

This work presents a novel approach for using acoustic feature selection as a decision-making technique and implements it using deep reinforcement learning. Specifically, we use a Deep-Q-Network (DQN) agent to navigate a high dimensional feature space of over 6,000 acoustic features extracted using the openSMILE toolkit, dynamically constructing maximally discriminative and non-redundant features subsets. In order to capture the latent structural dependencies among

acoustic features which classifier and wrapper methods have difficulty to model, we introduce the Graph Convolutional Network (GCN) based correlation awareness feature representation layer that operates as an auxiliary input to the DQN state encoder. Post selection interpretability is reinforced through TF-IDF weighting and K-means clustering which together yield both feature level and cluster level explanations that are clinically actionable. The framework is evaluated across five classifiers, namely, support vector machines (SVM), logistic regression, XGBoost, random forest, and feedforward neural network. We use 10-fold stratified cross-validation on established benchmarks of datasets, including DementiaBank Pitt Corpus, Ivanova, and ADReSS challenge data. The proposed approach is benchmarked against state-of-the-art feature selection methods such as LASSO, Recursive feature selection, and mutual information selectors. This research contributes to three primary intellectual advances: (1) a graph augmented state representation that encodes inter-feature relational structure within a reinforcement learning agent, (2) a clinically interpretable pipeline that bridges the gap between algorithmic performance and translational utility, and (3) multilingual data approach for the reinforcement learning agent framework. This study has direct implications for equitable, low-cost and scalable AD screening in both clinical and community settings.

Federated learning (FL) has emerged as a promising distributed machine learning
framework that enables multiple devices to collaboratively train models without sharing raw
data, thereby preserving privacy and reducing the need for centralized data collection. However,
deploying FL in practical wireless environments introduces two major challenges. First, the data
generated across distributed devices are often heterogeneous and non-IID, which makes a single
global model insufficient for many users. Second, learning performance in wireless systems is
strongly affected by communication constraints such as interference, unreliable channels, and
dynamic resource availability. This PhD research aims to address these challenges by bridging
FL methods and wireless networks.
In the first thrust, we develop personalized and adaptive FL methods given the underlying
wireless link conditions. To this end, we propose channel-aware neighbor selection and
similarity-aware aggregation in wireless device-to-device (D2D) learning environments. We
further investigate the impacts of partial model update reception on FL performance. The
overarching goal of the first thrust is to enhance FL performance under wireless constraints.
Next, we investigate the opposite direction and raise the question: How can FL-based distributed
optimization be used for the design of next-generation wireless systems? To this end, we
investigate communication-aware participation optimization in vehicular networks, where
wireless resource allocation affects the number of clients that can successfully contribute to FL.
We further extend this direction to integrated sensing and communication (ISAC) systems,
where personalized FL (PFL) is used to support distributed beamforming optimization with joint
sensing and communication objectives.
Overall, this research establishes a unified framework for bridging FL and wireless networks. As
a future direction, this work will be extended to more realistic ISAC settings with dynamic
spectrum access, where communication, sensing, scheduling, and learning performance must be
considered jointly.

Person re-identification (Re-ID) and multi-object tracking in unconstrained surveillance environments pose significant challenges within the field of computer vision. These complexities stem mainly from occlusion, variability in appearance, and identity switching across various camera views. This research outlines a comprehensive and innovative agenda aimed at tackling these issues, employing a series of increasingly advanced deep learning architectures, culminating in a groundbreaking occlusion-aware Vision Transformer framework.

At the heart of this work is the introduction of Deep SORT with Multiple Inputs (Deep SORT-MI), a cutting-edge real-time Re-ID system featuring a dual-metric association strategy. This strategy adeptly combines Mahalanobis distance for motion-based tracking with cosine similarity for appearance-based re-identification. As a result, this method significantly decreases identity switching compared to the baseline SORT algorithm on the MOT-16 benchmark, thereby establishing a robust foundation for metric learning in subsequent research.

Expanding on this foundation, a novel pose-estimation framework integrates 2D skeletal keypoint features extracted via OpenPose directly into the association pipeline. By capturing the spatial relationships among body joints along with appearance features, this system enhances robustness against posture variations and partial occlusion. Consequently, it achieves substantial reductions in false positives and identity switches compared to earlier methods, showcasing its practical viability.

Furthermore, a Diverse Detector Integration (DDI) study meticulously assessed the influence of detector choices—including YOLO v4, Faster R-CNN, MobileNet SSD v2, and Deep SORT—on the efficacy of metric learning-based tracking. The results reveal that YOLO v4 consistently delivers exceptional tracking accuracy on both the MOT-16 and MOT-17 datasets, establishing its superiority in this competitive landscape.

In conclusion, this body of research notably advances occlusion-aware person Re-ID by illustrating a clear progression from metric learning to pose-guided feature extraction and ultimately to transformer-based global attention modeling. The findings underscore that lightweight, meticulously parameterized Vision Transformers can achieve impressive generalization for occlusion detection, even under constrained data scenarios. This opens up exciting prospects for integrated detection, localization, and re-identification in real-world surveillance systems, promising to enhance their effectiveness and reliability.

Android malware remains a significant threat to mobile security, requiring efficient and scalable detection methods. This project presents an Android Malware Detection System that uses machine learning to classify applications as benign or malicious based on static permission-based analysis. The system is trained on the TUANDROMD dataset of 4,464 applications using four models-Logistic Regression, XGBoost, Random Forest, and Naive Bayes-with a 75/25 train/test split and 5-fold cross-validation on the training set for evaluation. To improve reliability, the system incorporates a hybrid decision approach that combines machine learning confidence scores with a rule-based static analysis engine, using a three-zone confidence routing mechanism to capture threats that ML alone may miss. The solution is deployed as a Flask web application with both a manual detection interface and an APK file scanner, providing predictions, confidence scores, and risk insights, ultimately supporting more informed and secure decision-making.

Architectural design space exploration (DSE) requires navigating large configuration spaces while satisfying multiple conflicting objectives, including performance, cost, and system dependability. Large language models (LLMs) have shown promise in assisting DSE by proposing candidate designs and interpreting simulation feedback. However, extending LLM-based DSE to realistic multi-objective settings introduces structural challenges. A naive multi-objective extension of prior LLM-based DSE approaches, which we term Co-Pilot2, exhibits reasoning instability, candidate degeneration, feasibility violations, and lack of progressive improvement. These limitations arise not from insufficient model capacity, but from the absence of structured control, verification, and decision integrity within the exploration process. 

To address these challenges, this research introduces REMODEL, a structured LLM-controlled DSE framework that transforms free-form reasoning into a constrained, verifiable, and iterative optimization process. REMODEL incorporates candidate pooling across parallel reasoning instances, strict state isolation via history snapshotting, deterministic feasibility verification, canonical design representation and deduplication, explicit decision stages, and structured reasoning to enforce complete parameter coverage and consistent trend analysis. These mechanisms enable reliable and stable exploration under complex multi-objective constraints. 

To support dependability-aware evaluation, the framework is integrated with cycle-accurate simulation using gem5 and its reliability-focused extension GemV, enabling detailed analysis of performance, power, and fault tolerance through vulnerability metrics. This integration allows the system to reason not only about performance–cost trade-offs, but also about reliability-aware design decisions under realistic execution conditions. 

Experimental evaluation demonstrates that REMODEL identifies near-optimal designs within a small number of simulations, achieving significantly higher solution quality per simulation compared to baseline methods such as random search and genetic algorithms, while maintaining low computational overhead. 

This work establishes a foundation for dependable LLM-assisted DSE by incorporating reliability constraints into the exploration loop. As a future direction, this framework will be extended to incorporate security-aware design considerations, enabling unified reasoning over performance, cost, reliability, and system security. 

This thesis addresses two related problems in the optical manipulation of microscopic particles: the efficient generation of holograms for holographic optical tweezers and the simulation of multi-particle optical binding. Holographic optical tweezers use phase-only spatial light modulators to create programmable optical trapping fields, enabling dynamic control over the number, position, and relative strength of optical traps. Because the quality of the trapping field depends strongly on the computed hologram, the first part of this work focuses on improving hologram-generation methods used in these systems.

A new phase-induced compressive sensing algorithm is presented for holographic optical tweezers, along with weighted and unweighted variants. These methods are developed from the Gerchberg-Saxton framework and are designed to improve computational efficiency while preserving favorable trapping characteristics such as uniformity and optical efficiency. By combining compressive sensing with phase induction, the proposed algorithms reduce the computational burden associated with iterative hologram generation while maintaining strong performance across a variety of trapping arrangements. Comparative simulations are used to evaluate these methods against several established hologram-generation algorithms, and the results show that the proposed approaches offer meaningful improvements in convergence behavior and overall performance.

The second part of this thesis examines optical binding, a phenomenon in which multiple particles interact through both the incident optical field and the fields scattered by neighboring particles. To study this process, a numerical simulation is developed that incorporates gradient forces, radiation pressure, and light-mediated particle-particle interactions in both two- and three-dimensional configurations. The simulation is used to investigate how particles evolve under different initial conditions and illumination states, and how collective effects influence the formation of stable or semi-stable arrangements. These results provide insight into the role of scattering-mediated forces in many-particle optical systems and highlight differences between two-dimensional and three-dimensional behavior.

Although hologram generation and optical binding are treated as separate problems in this work, they are connected by a common goal: understanding how structured optical fields can be designed and applied to control microscopic matter. Together, the results of this thesis contribute to the broader study of computational beam shaping and many-body optical interactions, with relevance to advanced optical trapping, particle organization, and dynamically reconfigurable light-driven systems.

Runtime estimation plays a structural role in reservation-based scheduling for High Performance Computing (HPC) systems, where predicted walltimes directly influence reservation timing, backfilling feasibility, and overall queue dynamics. This raises a fundamental question of whether improved runtime prediction accuracy necessarily translates into improved scheduling performance. In this work, we conduct an empirical study of runtime estimation under EASY Backfilling using an application-driven workload consisting of MRI-based brain segmentation jobs. Despite identical configurations and uniform metadata, runtimes exhibit substantial variability driven by intrinsic input structure. To capture this variability, we develop a feature-driven machine learning (ML) framework that extracts region-wise features from MRI volumes to predict job runtimes without relying on historical execution traces or scheduling metadata. We integrate these ML-derived predictions into an EASY Backfilling scheduler implemented in the Batsim simulation framework. Our results show that regression accuracy alone does not determine scheduling performance. Instead, scheduling performance depends strongly on estimation bias and its effect on reservation timing and runtime exceedances. In particular, mild multiplicative calibration of ML-based runtime estimates stabilizes scheduler behavior and yields consistently competitive performance across workload and system configurations. Comparable performance can also be observed with certain levels of uniform overestimation; however, calibrated ML predictions provide a systematic mechanism to control estimation bias without relying on arbitrary static inflation. In contrast, underestimation consistently leads to severe performance degradation and cascading job terminations. These findings highlight runtime estimation as a structural control input in backfilling-based HPC scheduling and demonstrate the importance of evaluating prediction models jointly with scheduling dynamics rather than through regression metrics alone.

The University of Kansas maintains hundreds of departmental and unit websites, leaving students without a unified way to find information. General-purpose chatbots hallucinate KU-specific facts, and static FAQ pages cannot hold a conversation. This work presents BabyJay, a Retrieval-Augmented Generation chatbot that answers student questions using content scraped from official KU sources, with inline citations on every response. The pipeline combines query preprocessing and decomposition, an intent classifier that routes most queries to fast JSON lookups, hybrid retrieval (BM25 and ChromaDB vector search merged via Reciprocal Rank Fusion), a cross-encoder re-ranker, and generation by Claude Sonnet 4.6 under a context-only system prompt. Evaluation on 46 question-answer pairs across five difficulty tiers and eight domains produced a composite score of 0.72, entity precision of 93%, and zero runtime errors. Retrieval, rather than generation, emerged as the primary bottleneck, motivating future work on multi-domain query handling.

Modern intelligent systems increasingly rely on continuous sensor data streams for perception, decision-making, and control, making sensors a critical yet underexplored attack surface. While prior research has demonstrated the feasibility of sensor-based attacks, recent advances in mobile operating systems and machine learning-based defenses have significantly reduced their practicality, rendering them more detectable, resource-intensive, and constrained by evolving permission and context-aware security models.

This dissertation revisits sensor exploitation under these modern constraints and develops a unified, cross-layer perspective that improves both practicality and stealth of sensor-enabled attacks. We identify three fundamental challenges: (i) the difficulty of reliably manipulating physical sensor signals in noisy, real-world environments; (ii) the effectiveness of context-aware defenses in detecting anomalous sensor behavior on mobile devices, and (iii) the lack of lightweight coordination for practical sensor-based side- and covert-channels.

To address the first challenge, we propose a physical-domain attack framework that integrates signal modeling, simulation-guided attack synthesis, and real-time adaptive targeting, enabling robust adversarial perturbations with high attack success rates even under environmental uncertainty. As a case study, we demonstrate an infrared laser-based adversarial example attack against face recognition systems, which achieves consistently high success rates across diverse conditions with practical execution overhead.

To improve attack stealth against context-aware defenses, we introduce an auto-contextualization mechanism that synchronizes malicious sensor actuation with legitimate application activity. By aligning injected signals with both statistical patterns and semantic context of benign behavior, the approach renders attacks indistinguishable from normal system operations and benign sensor usage. We validate this design using three Android logic bombs, showing that auto-contextualized triggers can evade both rule-based and learning-based detection mechanisms.

Finally, we extend sensor exploitation beyond the traditional attack-channel plane by introducing a lightweight control-plane protocol embedded within sensor data streams. This protocol encodes control signals directly into sensor observations and leverages simple signal-processing primitives to coordinate multi-stage attacks without relying on privileged APls or explicit inter-process communication. The resulting design enables low-overhead, stealthy coordination of cross-device side- and covert-channels.

Together, these contributions establish a new paradigm for sensor exploitation that spans physical, contextual, and control-plane dimensions. By bridging these layers, this dissertation demonstrates that sensor-based attacks remain not only feasible but also practical and stealthy in modern computer systems.