Computer Engineering Commons^™

Exploring Instruction Generation For Uavs: Dataset Adaptation, Model Behavior, And Diagnostic Insights, Seyedarman Vaziri Bozorg

Computer Science and Engineering Theses - Archive

This thesis explores the development of an answering agent capable of generating natural language instructions for unmanned aerial vehicles (UAVs), grounded in a limited, real-world dialogue dataset. The objective is to adapt a static dataset into a training pipeline that can support instruction generation and serve as a foundation for future interactive systems involving question-asking agents and internal dialogue. A hybrid architecture is implemented using a semantic teacher model (MPNet) and a T5-base encoder-decoder trained with contrastive and supervised objectives. The adapted training process yields statistically acceptable performance across standard evaluation metrics. However, qualitative analysis reveals a mismatch between metric …

Training Data Privacy In Machine Learning: A Systematization Of Attacks And Defenses, Mohammad Sufyaan Saeed

Computer Science and Engineering Theses - Archive

Training and deploying Machine Learning (ML) models introduce significant data confidentiality risks, as modern models can inadvertently memorize and leak information about their training data. While attacks such as membership inference and model inversion are well studied, the literature remains fragmented, with inconsistent threat models and unclear relationships across attack classes and defenses. This work presents a Systematization of Knowledge (SoK) that unifies the landscape of training-data privacy attacks and defenses, aligning them with the NIST Adversarial Machine Learning (AML) taxonomy to enable standardized threat modeling and comparison. Our analysis shows that, despite significant progress in characterizing attack vectors, defenses …

Transformer And Recurrent Architectures For Dynamics Prediction And Policy Learning On Long-Horizon Tasks, Vinal Jitendrabhai Gadhiya

Computer Science and Engineering Theses - Archive

Model-based reinforcement learning promises improved sample efficiency by learning environment dynamics and using them for planning or policy improvement. However, the choice of neural architecture for dynamics prediction significantly impacts the model's ability to capture temporal dependencies and maintain long-term context, capabilities crucial for complex, open-world environments.

This thesis investigates three neural architectures for learning world models: Transformer-based, GRU-based, and a hybrid Transformer+GRU approach. We evaluate these architectures on Crafter, a 2D open-world survival environment that requires long-horizon planning and sequential task completion. In Crafter, agents must perform hierarchical sequences of actions, such as collecting wood, placing a table, and …

Multi-Modal Model-Based Optical Flow Estimation For Event-Based Vision, Pritam Karmokar

Computer Science and Engineering Dissertations - Archive

Event cameras offer a fundamentally different sensing paradigm by asynchronously capturing brightness changes at high temporal resolution, directly encoding motion in the scene. However, their sparse and non-traditional data format poses significant challenges for dense motion estimation, particularly in the context of optical flow. Contrast Maximization (CM) has emerged as a powerful model-based framework for estimating optical flow from event data by optimizing the sharpness of motion-compensated event representations. This dissertation builds upon and significantly advances the CM framework through two complementary contributions.

First, we propose Edge-Informed Contrast Maximization (EINCM), a hybrid approach that augments the traditional events-only CM framework …

Deep Neural Network Models For Heatsink Performance Prediction And Optimization In Single Phase Immersion Cooling: Framework For Future Design Tools And Digital Twin Integration, Braxton J. Smith

Mechanical and Aerospace Engineering Theses - Archive

The rapidly rising computational power of modern computing components combined with the advanced packaging techniques being implemented has resulted in exponentially increasing thermal design powers (TDP) from CPUs and GPUs. Traditional air-cooling methods are approaching their effective cooling limits for many of these components, requiring lower supply air temperatures, higher supply air flowrates, and much larger heatsinks to remain feasible. Transitioning from air-cooling to single-phase immersion cooling offers numerous benefits in thermal performance, data-center size reduction, and energy efficiency. To leverage the merits of immersion cooling, the performance of a given heatsink must be predicted and optimized for best performance …

Medical Ai: Solving Healthcare Challenges And Inspiring Ai Innovation, Xiaowei Yu

Computer Science and Engineering Dissertations - Archive

Artificial Intelligence (AI) is transforming healthcare by enabling large-scale analysis of medical data and integrating multimodal information for more comprehensive diagnostics. I present my work addressing fundamental and challenging problems in developing state-of-the-art AI models for medical data analysis, including multimodal brain data and other medical datasets. Additionally, I design brain-inspired AI models by integrating insights from organizational principles of brain networks. Specifically, my research tackles three critical aspects: (1) AI in Computational Neuroscience, where I design deep learning models for brain network analysis to uncover the organizational principles of brain networks; (2) Brain-Inspired AI, where I integrate superior brain …

Immersive Executive Functions Assessment System (Iexec): Integrating Embodied Cognition And Virtual Reality, Hamza Reza Pavel

Computer Science and Engineering Dissertations - Archive

Executive functions (EFs) are higher-order cognitive processes that include working memory, inhibitory control, and cognitive flexibility. These higher-order processes facilitate the achievement of goal-directed behavior and enable both adaptive decision-making and emotional regulation. Traditional EF assessment tools depend on static pen-and-paper tasks or basic computer-based tasks, which fail to capture real-world cognitive complexity and dynamics. Some of these assessment tools are specifically geared towards children or older adults, while others are more generic and designed to be used for people of all ages. This dissertation addresses these limitations by introducing iExec: The Immersive Executive Functions Assessment System, which functions as …

Enabling Energy And Water Sustainability Through Out-Of-Band Emi Sensing And Infrastructure Modeling, Pranjol Sen Gupta

Computer Science and Engineering Dissertations - Archive

As demand for Internet and cloud services surges, data centers have emerged as critical infrastructure—but they are also among theworld’s most energy- andwater-intensive facilities. Effective power management, particularly at the server level, is essential for improving efficiency, reliability, and sustainability. However, server-level power monitoring remains uncommon due to the high cost of hardware instrumentation and the intrusiveness of software-based solutions, especially in shared colocation environments. My research introduces a novel, low-cost, and non-intrusive method for server-level power monitoring using conducted electromagnetic interference (EMI). By analyzing EMI signals captured from higher levels in the power distribution network, this approach estimates individual …

Toward Strategy Identification And Subtask Decomposition In Task Exploration, Tom Odem

Master's Projects

This research builds on work in anticipatory human-machine interaction, a subfield of human-machine interaction where machines can facilitate advantageous interactions by anticipating a user’s future state. The aim of this research is to further a machine’s understanding of user knowledge, skill, and behavior in pursuit of implicit coordination. A task explorer pipeline was developed that uses clustering techniques, paired with factor analysis and string edit distance, to automatically identify key global and local strategies that are used to complete tasks. Global strategies identify generalized sets of actions used to complete tasks, while local strategies identify sequences that used those sets …

Faux Capabilities: A Novel Approach For Code Analysis, Tanay Godse

Master's Projects

When using third-party packages or libraries, it is crucial to understand their behavior. Typically, this requires developers to either conduct code reviews or set up sandbox environments for testing or write unit tests with mocked values for every function used in their code. However, these approaches are often inefficient and time-consuming. A more effective solution would provide developers with a broad understanding of the functionality required by the code they plan to import. This can be done using object capabilities, where a particular functionality is the capability that an object must possess, in order to be able to perform the …

Disease Diagnosis Using Rag Llm With Smart Prompt Engineering, Qadeerullah Syed

Master's Projects

Although recent trends indicate that LLMs outperform traditional methods in solving complex problems with enhanced reasoning, there has been barely any progress in replicating the quality of diagnoses like those of actual human doctors. The identification of an accurate diagnosis with thorough reasoning is still a significant challenge, even with advanced AI models. The process of performing accurate diagnosis remains challenging due to a lack of transparency in state-of-the-art models existing today, a lack of explanation in the diagnosis process, an emphasis on results rather than reasoning, and a lack of foundational knowledge in models, along with limited exploration of …

Malware Generation And Classification Using Pixelcnn, Mounika Krishna Teja Karumudi

Master's Projects

Malware poses a serious threat to both data privacy and system security. With the wide variety of malware families and the surge in cyber-attacks, the accurate classification of malware is crucial for building effective detection and prevention systems. In recent years, deep learning (DL) methods in computer vision have shown promise in classifying malware by converting malware files into visual representations and applying DL algorithms to classify the resulting images. Among the different approaches to malware family classification, image-based methods have gained significant interest. This research focuses on leveraging DL techniques for image-based classification of malware. The success of identifying …

Smart: Semantic Mapping And Analysis For Regional Terrain Using Multi-Scale U-Net And Topsis, Rashmi Sonth

Master's Projects

Accurate land use classification is the backbone for urban planning. But with poor quality satellite images, varied landscapes and structures which are changing faster than ever, it becomes a challenge to define clear boundaries and hence to urban planning. This research explores the application of deep-learning model for land use classification and asses the suitability of the land. The proposed model combines a multi-scale U-Net architecture with Transformer blocks applied on a multi-spectral satellite images that improves the semantic segmentation greatly across the urban and rural regions. Additionally, a patch-wise segmentation is applied to overcome the common problem of feature …

Multimodal Feature Fusion And Machine Learning For Adhd Detection Using Neuroimaging Data, Isabel Pham

Master's Projects

Attention Deficit Hyperactivity Disorder (ADHD) is a common neurodevelopment disorder that can significantly affect a person’s attention, impulse control, and executive function. Currently, the traditional diagnosis method often relies on clinical assessments and observations. However, these methods can be subjective and lead to inconsistencies in diagnosis between individuals. To address this challenge, neuroimaging and machine learning (ML) are promising tools for providing a more objective diagnosis of ADHD. The goal of this project is to apply a multimodal approach in which structural and functional features of specific regions of the brain are used to develop a more accurate and objective …

Simple Vs. Complex Human Activity Classification Via Hybrid Machine Learning Models, Anusha Kukreja

Master's Projects

In-depth understanding of the complexity of daily human activities is crucial for building responsive health monitoring and assistive technologies. However, limited research has focused on distinguishing activities based on their involvement level, as most existing work classifies only the type of activity performed. In this thesis, we address this gap by proposing a method to classify human activities as either simple or complex using sensor data from the Opportunity dataset. We define complex activities as those involving object interactions or multiple coordinated movements (e.g., drinking from a cup, cleaning a table), and simple activities as static or low-effort postures (e.g., …

Visionmate: Ai-Powered Image Captioning Web Application, Sai Anoushka Kokku

Master's Projects

VisionMate is a web application that generates captions for camera-captured images. It is designed
to assist users with visual impairments by converting visual input into spoken and written text. The
application uses the GIT-base model from Hugging Face, which processes the image and returns a
descriptive caption. Users can take a picture using the device camera—either via webcam on
desktop or the native camera interface on mobile. The app provides audio output using the
SpeechSynthesis API and uses full-screen tap interaction to simplify accessibility.

The frontend is implemented in React.js, and the backend is built with FastAPI. The backend calls …

Phishing Detection Using Continual Learning And Large Language Models, Gopi Prajeev Battula

Master's Projects

Adaptive phishing detection remains crucial as the nature of cyber-attacks changes over time, which renders static models obsolete. This project extends phishing detection through the implementation of continual learning approaches, namely Elastic Weight Consolidation (EWC) and Learning Without Forgetting (LWF) with RoBERTa, a Large Language Model (LLM) and compares the results of these approaches against GPT-4o-mini, another LLM. Our approach begins with fine-tuning RoBERTa on multiple phishing datasets to establish an effective baseline. EWC is then implemented to preserve vital model parameters based on their importance measured by the Fisher Information Matrix, while LWF uses knowledge distillation to retain prior …

On The Adversarial Robustness Of Quantized Neural Networks Against Common Adversaries In Time-Series Forecasting, Maanak Arora

Master's Projects

Real-world edge applications now use modern machine learning models which require both resource efficiency and robustness against adversarial threats. Deep neural networks which include time series forecasting models still face risks from adversarial perturbations while quantization techniques used for memory and compute efficiency create unpredictable robustness challenges. This project investigates the adversarial resistance of Long Short-Term Memory (LSTM) models after applying post-training quantization at three different precision levels: 16-bit floating point (FP16), 8-bit integer (INT8) and custom 4-bit quantization. The Jena Climate dataset serves as our main benchmark for training a fullprecision LSTM model followed by multiple quantization strategies which …

Physiotrack: A Gamified Physiotherapy System, Pranavi Chaturvedula

Master's Projects

Traditional physiotherapy methods tend to be non-interactive and provide little to no personalized instruction, even though physiotherapy is critical to stroke recovery. This thesis explores a fully adaptive, sensor-based, feedback architecture intended for stroke patients which remotely supervises movement and personalizes exercises enabled by multimodal sensors. The system uses filtering and windowed segmentation of accelerometer and skeletal data to compute features like jerk, speed, and joint movement angular range. A game engine applies accelerometer and skeletal features together with optimized, lightweight ML models to drive adaptive feedback, scoring, and difficulty adjustment. The architecture supports responsive continuous sensor streaming within the …

Secured Data Storage Management With Deduplication In Cloud Computing And Local Gpt Integration, Pavan Myana

Master's Projects

Exponential growth in cloud computing has brought enormous changes in data storage and processing, but also raised several questions on the security, privacy, and efficient storage of data. This report provides a dual-focused approach toward solving these challenges. First, we try to build an application securely and efficiently using data deduplication and Proxy Re-Encryption for optimization of storage and enabling secure data sharing. Deduplication ensures that redundant data is removed before encryption for maximum efficiency in storage, while PRE enables the safe sharing of encrypted data by re-encrypting the keys for specified recipients without the leakage of sensitive information. We …

Performance Comparison Of Machine Learning Across Metal, Cuda, And Neuromorphic Frameworks, Ryan Saini

Master's Projects

Machine learning’s computational demands necessitate optimal performance and utilization. This research compares Apple Silicon M3 Pro with MPS, NVIDIA RTX 3070 GPU with CUDA, and neuromorphic computing for machine learning methods. We provide a cross-platform and cross-architecture performance analysis of machine learning methods to identify optimal configurations for training and inference scenarios. On traditional neural networks, Apple Silicon with MPS delivers superior energy efficiency at the cost of longer processing times for training and inference. NVIDIA with CUDA offers faster computation in training and inference at higher energy costs. Convolutional spiking neural networks perform competitively on event-based data, particularly on …

Malware Opcode Embedding And Quality Assessment Of Generative Sample Embeddings, Atishay Jain

Master's Projects

Malware is software used to damage and disrupt computer systems with the intent to cause damage to the victim. Malware detection and classification into malware families is a crucial problem for cybersecurity researchers. One of the major bottlenecks in improving these systems is the shortage of good quality labeled malware data, especially for malware families with scarce samples. Researchers have utilized generative models to generate malware data to address this issue. Malware embeddings encode patterns within a malware file, which can be used to detect and classify malware. Recently, encouraging results have been obtained in generating malware embeddings using generative …

Gen Ai For Malicious Network Data, Aneesh Maturu

Master's Projects

Though botnet attacks are on the rise, they also have become sophisticated and difficult to detect. Such a rising threat demands more and more sophisticated cybersecurity that leverages machine learning technology. Nevertheless, one of the biggest bottlenecks remains the unavailability of large and well-balanced datasets, particularly for malicious traffic, which hampers the efficacy of detection models. In an attempt to address this issue, our research utilizes Generative Adversarial Networks (GANs) to produce synthetic samples of botnet traffic from the CTU-13 dataset. While the majority of generative models have been targeting image data, we use GANs for a new application: generating …

Human Vs. Llm: Interpreting Emotion From Neutral Faces In Contextually Charged Backgrounds, Sai Mounika Peteti

Master's Projects

In this study, we investigate the intersection of cultural context, visual perception, and implicit bias in interpreting neutral human expressions. Specifically, we explore how emotionally charged backgrounds can shape viewers interpretations of neutral facial expressions. Using an experimental setup, participants are shown neutral human portraits paired with varying background types ranging from emotionally neutral or pleasant to evocative scenes. Participants are tasked with selecting which background best matches the emotional state of the person depicted in the portraits. Eye-tracking data is collected to analyze visual attention patterns and cognitive processing. In parallel, we also evaluate how large language models (LLMs) …

Immersive Extended Reality For Lower Limb Rehabilitation: Design, Deployment, And Pilot Study, Jeremy Varghese

Electronic Theses & Dissertations (2024 - present)

This thesis presents the design, deployment, and pilot study of an immersive extended-reality (XR) rehabilitation system integrated with a ceiling-mounted dynamic body-weight support device (Vector Gait and Safety System), aimed at improving lower-limb rehabilitation out- comes. The implemented system combined immersive virtual tasks—such as Touch Wall, Ball Launcher, Obstacle Dodge, and Stepping Stones—with real-time movement tracking, enabling detailed kinematic analysis and personalized therapy. A pilot study conducted at Sunnyview Rehabilitation Hospital involved seven patients with various mobility impairments, providing quantitative performance metrics and quali- tative user feedback. Results demonstrated consistent patient engagement, measurable im- provements in gait speed and task …

Introducing Catalizer: A Framework For Prototyping Models Of Biological Systems, Andy M. Day

Honors Theses

Modeling the light response system of Nannochloropsis oceanica brings

a set of challenges that make modeling difficult. Notably, potential models

may contain a large number of chemical species. A large number of

species creates a quadratic explosion in the number of potential pathways.

In addition, mathematically defining these pathways is error prone, yet

follows a surprising simple set of rules. We seek to create a domain specific

language which can precisely define these chemical reaction networks.

Once the networks have been defined, they can be exported as procedures

defined in popular programming languages for further analysis.

Evolving Secure Authentication From 5g To 6g: Advancing Privacy And Resilience In Next-Generation Networks, Isabella Deanne Lutz

Graduate Theses, Dissertations, and Problem Reports (ETD)

The fifth generation (5G) of mobile networks introduced groundbreaking improvements in connectivity, latency, and reliability. As 5G continues to expand across commercial and de- fense sectors, ensuring the privacy and integrity of its authentication mechanisms remains paramount. The foundation of 5G security lies in the Authentication and Key Agreement (AKA) protocol, which enhances user identity protection and establishes mutual authenti- cation between the user equipment (UE) and the network. Despite these advances, several weaknesses persist, including replay-based desynchronization, linkability, and correlation at- tacks under realistic adversary models. This thesis provides a unified analysis of these vulnerabilities and introduces a lightweight …

Implementation Of A Neural Network Execution Framework For Generalized And Cross-Platform Deep Learning Deployment And Inference On Spacecraft Systems, Rafael Polanco Segovia

Graduate Theses, Dissertations, and Problem Reports (ETD)

Recent advances in hardware and software technology have made it possible to implement more resource-demanding deep learning algorithms in constrained hardware environments. This creates opportunities to use deep learning for aerospace applications on increasingly smaller aerospace vehicles. This work presents the implementation of a Neural Network Execution Framework (NNEF), which aims to provide a cross-platform and reusable framework to deploy and execute trained neural networks for deep learning aerospace applications. The NNEF executes any neural network inference process regardless of the original deep learning framework in which it was created, for supported flight software platforms, and space-like computer boards. Users …

Hypnosis And Mindfulness Audio Recordings For Reducing Fatigue In Individuals With Multiple Sclerosis: A Randomized Controlled Study, Mark P. Jensen, Susan Robles, Michael G. Nash, Susanne May, Dwan M. Ehde, Melissa A. Day, Owen Gottlieb, Laurence I. Sugarman, Kevin N. Alschuler

Articles

Background

Fatigue is a common problem in individuals with multiple sclerosis (MS).

Objective

The objective was to evaluate the effects on fatigue of having 4 weeks of

access to audio recordings of therapeutic hypnosis (HYP) and mindfulness meditation

(MM) practices.

Methods

A total of 333 individuals with MS and fatigue were randomly assigned to

one of the three treatment conditions for 28 weeks: (1) access to therapeutic HYP audio

recordings, (2) access to MM audio recordings, or (3) no access to recordings

(treatment as usual or TAU). Fatigue impact (primary outcome) and other outcomes

were assessed at 4, 16, and …

Earthquake Wrangler: Leveraging Ios Technology For Earthquake Detection And Early Warning Application To Enhance Public Safety., Luis F. Salome

Theses and Dissertations--Civil Engineering

Earthquakes are devastating natural phenomena and generate secondary hazards such as tsunamis, landslides and fires. Their catastrophic impacts span both developed nations including the United States, Japan, Turkey, and Italy and developing countries such as El Salvador, Haiti, Nepal and the Philippines, where disparities in early warning infrastructure remain important. Seismic events start with stress waves generated by tectonic plate motion, with body waves (P- waves and S-waves) and surface Rayleigh and love waves that carry energy through the earth. While some regions have adopted advanced early warning systems based on seismic hazard models and strong ground motion analysis, others …