Digital Commons Network^™

Grm-1249 Ai-Assisted Diabetic Retinopathy Screening From Fundus Images, Mohan Krishna Thiriveedhi, Tarun Teja Pokala

C-Day Computing Showcase

Diabetic Retinopathy (DR) is a major cause of avoidable blindness among diabetic patients worldwide. Early screening is critical, but manual diagnosis is time-consuming and requires specialists. This paper presents a deep learning system to automatically analyze retinal fundus images and perform a focused, binary classification to distinguish between 'No DR' (Healthy) and 'Severe-Stage DR' (Severe/Proliferative). We benchmark three prominent architectures: a ResNet-50, an EfficientNet-B0, and a Vision Transformer (ViT-B/16). The models are trained and evaluated on a custom-balanced, binary dataset derived from the APTOS 2019 collection. We conduct two experiments, one with a small dataset (N=500) and one with a …

Uc-1244 Agentic Ai For Intelligent Customer Communication, Lucas Papadopoulos, Jeremy Hopkins, Munir Gargour, Weston Dease

C-Day Computing Showcase

E-commerce web shoppers need fast, reliable responses to a variety of requests: account modifications, order tracking, or policy inquiries. Businesses must address user queries in a fast and efficient manner, or else lose customers. Multi-agent AI models boast the ability to answer customer questions and act upon consumer queries without outside intervention. However, research is sparse as to how agentic models can transfer benefit to large commercial software stacks under realistic commercial load. We sought to ask whether a multi-agent AI architecture can effectively handle commercial-scale e-commerce customer service tasks. Moreover, we investigated how a multi-agent AI architecture compares to …

Uc-1175 Admissions Assistant - Ai Chatbot, Chelsea Oduro, Jeffery Bond, Tanner Morgan, Allen Scott, Joy Parks

C-Day Computing Showcase

The Kennesaw State University (KSU) Graduate Admissions website contains extensive information on programs, application processes, deadlines, and eligibility requirements. However, its dense structure can make it difficult for prospective students to quickly locate specific details. This often leads to repeated inquiries from admissions staff, increasing their workload and delaying responses to more complex applicant needs. This project addresses those challenges by introducing a chatbot capable of retrieving and presenting the website information in a more intuitive, conversational format, improving user satisfaction and reducing workflow overhead for admissions staff.

Uc-1232 Stronghold, Caitlin Tigani, Adam Tucker, Camden Lloyd, Dale Balsor

C-Day Computing Showcase

Our goal for this game is to create a game using both a single agent and multi agent AIs. In our game you can either play against the current AI or train the AI up as it fights against another AI. The training will use genetic AI. So whichever AI wins that row will be the one that move on. The loser will have their weights adjusted. This means the more that you train the Stronghold AI the harder the AI will be to fight against. The final game mode available is being able to play against another.

Uc-1234 Fundoria, Connor Stabler, Dameon Jones, Rafay Hassan, Michael Liu, Brenden Horne

C-Day Computing Showcase

Fundoria is an educational technology project designed to transform financial literacy and career readiness for elementary and middle-grade students. The platform combines interactive gameplay, AI-driven personalized learning, and real-world job simulations to teach essential financial concepts in an engaging, age-appropriate way. Students explore budgeting, saving, banking, and career skills through guided missions that adapt to their learning styles. Fundoria also provides teachers and schools with standards-aligned activities, classroom resources, and assessment tools that connect academic content to practical, everyday decisions. By blending gamification with evidence-based instruction, Fundoria empowers students to build confidence, develop responsible financial habits, and prepare for future …

Grm-1245 A Synthetic Data Engine For Explainable Injection-Area Perception, Yukang Shen

C-Day Computing Showcase

Vision-Language-Action (VLA) systems are beginning to support everyday clinical workflows. Deltoid intramuscular injection is a representative task, but progress is limited by data scarcity, privacy constraints, and the cost of expert annotation. Recent text-to-image (T2I) models make large-scale data synthesis possible, yet ensuring anatomical correctness, diversity, and label quality remains difficult. To address this gap, we propose a Synthetic Data Engine tailored for medical perception, integrating cold-start filtering, controlled T2I generation, CLIP-based quality checks, and iterative segmentation training. We further introduce an anthropometry-grounded formulation of injection safety that produces interpretable safe-zone guidance. Experiments show that synthetic data can effectively bootstrap …

Deconstructing The Black Box: An Explainability Analysis Of Deep Learning Architectures In Cytopathology, Chase A. Garrett

Master's Theses or Doctor of Nursing Practice

Deep learning shows strong potential in medical-image analysis, yet adoption in cyptopathology remains limited. Cytopathology could benefit from deep learning applications by improving diagnostic efficiency and accuracy. However deep learning comes with a notorious “black box” that keeps the models from being transparent and trustworthy for widespread clinical adoption. We conducted a comprehensive and comparative analysis of several deep learning architectures for multi-class classification of acute leukemia types, ALL, AML, and normal healthy cells from peripheral blood smear images. The models in this research include a Vision Transformer (ViT) and a diverse selection of Convolutional Neural Network (CNN) models. The …

Privacy In Flux: A 35-Year Review Of Trends, Legal Evolution, And Emerging Challenges, Kong Phang, Jihene Kaabi

Research & Publications

Privacy harms have expanded alongside rapid technological change, challenging the adequacy of existing regulatory frameworks. This systematic review (1990–2025) systematically maps documented privacy harms to specific legal mechanisms and observed enforcement outcomes across jurisdictions, using PRISMA-guided methods and ROBIS risk-of-bias assessment. We synthesize evidence on major regimes (e.g., GDPR, COPPA, CCPA, HIPAA, GLBA) and conduct comparative legal analysis across the U.S., E.U., and underexplored regions in Asia, Latin America, and Africa. Key findings indicate increased recognition of data subject rights, persistent gaps in cross-border data governance, and emerging risks from AI/ML/LLMs, IoT, and blockchain, including data breaches, algorithmic discrimination, and …

Archaeological Predictive Meta-Modeling In Pre-Columbian Mexico, Peter Stamm

Electronic Theses and Dissertations

Archaeological Predictive Modeling stands firmly as an important tool for Archaeologists to predict undiscovered sites from civilizations all across the globe. While powerful, this methodology is not without its own set of qualms. Striking a balance between pure a data-driven approach while also observing leading expert theories can be a complicated task. Going further, deciding on the specific domain of features to emphasize or overlook can be a challenge within itself, as one misstep can drastically change the output of model, sometimes for the worst. In addition, creating models that can expose their reasoning process can be rather difficult to …

From Vulnerability To Robustness: A Survey Of Patch Attacks And Defenses In Computer Vision, Xinyun Liu, Ronghua Xu

Michigan Tech Publications

Adversarial patch attacks have emerged as a powerful and practical threat to machine learning models in vision-based tasks. Unlike traditional perturbation-based adversarial attacks, which often require imperceptible changes to the entire input, patch attacks introduce localized and visible modifications that can consistently mislead deep neural networks across varying conditions. Their physical realizability makes them particularly concerning for real-world security-critical applications. In response, a growing body of research has proposed diverse defense strategies, including input preprocessing, robust model training, detection-based approaches, and certified defense mechanisms. In this paper, we provide a comprehensive review of patch-based adversarial attacks and corresponding defense techniques. …

Enhancing Recommendation Performance Via Stacking Ensemble And Synthetic Data Augmentation, Zahraa Yaareb Hani, Mohsin Hasan Hussein

Karbala International Journal of Modern Science

Recommendation systems are essential tools that primarily aim to help users navigate through a large volume of information. They simplify the decision-making process by suggesting relevant items based on users’ historical behaviour. However, their performance is often affected by common problems such as data sparsity. This work proposes a stacking-based ensemble recommendation system that integrates multiple machine learning models to enhance the model’s predictive performance. A new synthetic data augmentation technique is introduced to address the sparsity issue in the user–item rating matrix. This method uses the Naïve Bayes algorithm to predict additional ratings for each user. These are then …

Creating An Iot User Centric Security And Privacy Label, Luke Joseph Gleba

Theses and Dissertations

Concerns for cybersecurity awareness and training in the public have been rising at astronomical rates over the past few years. People globally have entered a critical intersection with computers. Computers are being used in everyday life, and users do not have an easy way to understand their own cyber risk. Consumers deserve to know how secure their new Internet of Things (IoT) system is in a quick, efficient way before they purchase the device and while they own it. The following research looks to improve on ideas for the criteria and the development of a security label. Few prototypes currently …

Mimar-Net: Multiscale Inception-Based Manhattan Attention Residual Network And Its Application To Underwater Image Super-Resolution, Nusrat Zahan, Sidike Paheding, Ashraf Saleem, Timothy C. Havens, Peter C. Esselman

Michigan Tech Publications

In recent years, Single-Image Super-Resolution (SISR) has gained significant attention in the geoscience and remote sensing community for its potential to improve the resolution of low-quality underwater imagery. This paper introduces MIMAR-Net (Multiscale Inception-based Manhattan Attention Residual Network), a new deep learning architecture designed to increase the spatial resolution of input color images. MIMAR-Net integrates a multiscale inception module, cascaded residue learning, and advanced attention mechanisms, such as the MaSA layer, to capture both local and global contextual information effectively. By utilizing multiscale processing and advanced attention strategies, MIMAR-Net allows us to handle the complexities of underwater environments with precision …

Game Hacking & Anti-Cheat Analysis, Quang Hoang

Cybersecurity Undergraduate Research Showcase

Reverse engineering and analyzing game hacking and anti-cheat mechanisms is a complex and evolving field. This research document explores the history of game hacking, various techniques used in game hacking, and the countermeasures implemented by anti-cheat systems. Through case studies, we illustrate the strategies involved in creating cheats. Specifically, we demonstrate how to hack the open-source game AssaultCube using memory editing and code injection techniques available in Cheat Engine in a step-by-step manner so that the reader can theoretically reproduce the results shown in this paper. We also dissect the game’s anti-cheat mechanisms, identifying their strengths and weaknesses. This document …

Implementation And Assessment Of The Openbci Platform As An Accessible Brain- Computer Interface, Jewell Norris

Honors Theses

OpenBCI is a low-cost, open-source platform for alternative brain-computer interface (BCI) software and hardware. This thesis evaluates OpenBCI’s electroencephalogram (EEG) and electromyography (EMG) capabilities by constructing and testing a 16-channel EEG system using the Ultracortex Mark IV headset and Cyton + Daisy biosensing board. The viability of the system was assessed through real-time BCI control and comparison to a clinical-grade EEG system. Real-time BCI control of an online falling-block game was tested via the use of eye blinks EMG (channels Fp1/Fp2) and head-tilt accelerometer inputs. The BCI game demonstrated reliable control despite minor latency and artifact sensitivity. For clinical comparison, …

Evolutionary Reinforcement Learning Based On Elite Instruction And Random Search, Jian Di, Xue Wan, Limei Jiang

Journal of System Simulation

Abstract: Evolutionary reinforcement learning currently suffers from low sample efficiency, a single coupling method, and poor convergence, which can affect its performance and scaling. To address this issue, an improved algorithm based on elite gradient instruction and double random search was proposed. The direction of the reinforcement strategy gradient update was corrected by introducing elite strategy gradient guidance carrying evolutionary information during reinforcement strategy training. Double stochastic search was used to replace the original evolutionary component, reducing the complexity of the algorithm while making the policy search meaningful and controllable in the parameter space. The introduction of complete replacement information …

Research On Time Sequence Design Method Of Dynamic Simulation Scene For Starlight Navigation, Xiaoting Su, Xiaowei Zhang, Yi Tian, Qi Li, Shuaihao Wang

Journal of System Simulation

Abstract: To solve the problem of misidentification of star maps due to time sequence mismatch in the hardware-in-the-loop simulation system of star navigation, where star trackers with different shutter types (global shutter and rolling shutter) and star simulators with varying refresh display methods (whole frame refresh and line sweep refresh) operated without synchronization, a time sequence design method of the dynamic simulation scene for starlight navigation without the need of external synchronization signals was proposed. The method could design the refresh frequency and duty cycle of the corresponding star simulators according to the detector integration time of the tested star …

Quality Assessment Of Pathology Board-Exam-Style Mcqs Produced By Chatgpt3.5: A Comparative Study, Arianna B. Morton, Zunaira Naeem, Allison F. Goldberg, Alexis R. Peedin, Joanna Chan

Department of Pathology, Anatomy, and Cell Biology Faculty Papers

Residents preparing for pathology board exams frequently use multiple-choice questions (MCQs) from question banks (QBs) like PathDojo and PathPrimer, which can be costly. ChatGPT, a free tool, has been used to generate MCQs for other tests like the SAT. This study compared the quality of pathology MCQs created by ChatGPT versus commercially available study questions for the American Board of Pathology’s (ABPath) certifying exams. A rubric adapted from the National Board of Medical Examiners’ (NBME) question writing guide was validated by two pathologists using commercially available pathology board exam questions. This rubric was then used to evaluate MCQs from commercially …

Speculative Automated Refactoring Of Imperative Deep Learning Programs To Graph Execution, Raffi Khatchadourian Ph.D., Tatiana Castro Vélez, Mehdi Bagherzadeh, Nan Jia, Anita Raja

Publications and Research

Efficiency is essential to support ever-growing datasets, especially for Deep Learning (DL) systems. DL frameworks have traditionally embraced deferred execution-style DL code---supporting symbolic, graph-based Deep Neural Network (DNN) computation. While scalable, such development is error-prone, non-intuitive, and difficult to debug. Consequently, more natural, imperative DL frameworks encouraging eager execution have emerged but at the expense of run-time performance. Though hybrid approaches aim for the "best of both worlds," using them effectively requires subtle considerations. Our key insight is that, while DL programs typically execute sequentially, hybridizing imperative DL code resembles parallelizing sequential code in traditional systems. Inspired by this, we …

Speculative Automated Refactoring Of Imperative Deep Learning Programs To Graph Execution, Raffi Khatchadourian Ph.D., Tatiana Castro Vélez, Mehdi Bagherzadeh, Nan Jia, Anita Raja

Publications and Research

Efficiency is essential to support ever-growing datasets, especially for Deep Learning (DL) systems. DL frameworks have traditionally embraced deferred execution-style DL code---supporting symbolic, graph-based Deep Neural Network (DNN) computation. While scalable, such development is error-prone, non-intuitive, and difficult to debug. Consequently, more natural, imperative DL frameworks encouraging eager execution have emerged but at the expense of run-time performance. Though hybrid approaches aim for the "best of both worlds," using them effectively requires subtle considerations. Our key insight is that, while DL programs typically execute sequentially, hybridizing imperative DL code resembles parallelizing sequential code in traditional systems. Inspired by this, we …

Secret Key Generation Based On The Physical Layer Characteristics For Iot Networks, Abdullah Dakhlallah Alshamdayn

Doctoral Dissertations

The rapid expansion of low-resource devices, coupled with advances in telecommunications, has significantly increased the number of connected devices and enabled the development of affordable, energy-efficient, portable, and high-performance sensors for diverse applications. However, this convenience comes with security and privacy concerns related to the reliability of hardware, software, and communication infrastructure. The extensive interconnectivity of limited-resource devices and the transmission of large data volumes pose significant security challenges in wireless networks. The future wireless technologies, such as 5G, will enable the transfer of critical data, including personal, financial, military, and industrial information, necessitating secure communication in wireless networks. Generally, …

Inverse Design For Generating Initial Conditions In Scientific Simulations, Leslie Horace, Christin Whitton, Vanessa Job, William Jones, Nathan A. Debardeleben

Computing Sciences

We propose a conditional normalizing flow (CNF) surrogate model to solve generative, many-to-one inverse problems in scientific simulations governed by partial differential equations (PDEs) with time-evolving interactions between heterogeneous materials. We present two case studies: electrostatic potential and heat diffusion, which serve as proxy simulations for generating diverse sets of initial conditions that can reproduce an observed output state (transient or steady). Finally, we provide a comprehensive overview of the synthetic datasets, the model specification, each stage of the experimental workflow, evaluation of training performance, and uncertainty quantification for the generated samples.

Modeling And Optimizing Real-Time Telescope Interaction For Multi-Wavelength Observation Of Gamma-Ray Bursts, Ye Htet, Marion Sudvarg, Honghao Yang, Jeremy Buhler, Roger Chamberlain, James Buckley

Computer Science Faculty Research & Creative Works

Multi-wavelength observation of gamma-ray bursts (GRBs) requires real-time interaction among multiple telescopes. A gamma-ray telescope detects and localizes a GRB in the sky and must then communicate with an optical telescope to direct the latter toward the GRB as quickly as possible. We previously developed software for ADAPT, a suborbital gamma-ray telescope, to localize GRBs in real time, on a timescale shorter than that of the GRB itself. This work therefore studies progressive localization, in which ADAPT computes a series of increasingly accurate location estimates during a GRB to enable a partner instrument to more rapidly find it. We describe …

Connected-Component Labeling Using Hls For High-Energy Particle Physics Instruments, Nick Song, Marion Sudvarg, Roger Chamberlain

Computer Science Faculty Research & Creative Works

Many instruments used in high-energy particle physics observations, e.g., gamma-ray telescopes, use FPGAs for front-end signal processing of raw sensor data. The use of high-level synthesis (HLS) to express the signal processing algorithms has the potential to significantly reduce development time for new instruments of this type. We describe our experience with one of the computational stages in the signal processing pipeline, island detection, exploring its implementation across multiple configurations: 1D versus 2D islands, and 4-way versus 8-way connected-component labeling (CCL) in the 2D configuration. We report resource usage and performance for both configurations of 2D island detection, including the …

Chronosort: Revealing Hidden Dynamics In Alphafold3 Structure Predictions, Matthew J. Argyle, William P. Heaps, Corbyn Kubalek, Spencer Gardiner, Bradley C. Bundy, Dennis Della Corte

Faculty Publications

Protein function emerges from dynamic conformational changes, yet structure prediction methods provide only static snapshots. While AlphaFold3 (AF3) predicts protein structures, the potential for extracting dynamic information from its ensemble predictions has remained underexplored. Here, we demonstrate that AF3 structural ensembles contain substantial dynamic information that correlates remarkably well with molecular dynamics simulations (MD). We developed ChronoSort, a novel algorithm that organizes static structure predictions into temporally coherent trajectories by minimizing structural differences between neighboring frames. Through systematic analysis of four diverse protein targets, we show that root-mean-square fluctuations derived from AF3 ensembles can correlate strongly with those from MD …

Deconstructing Tycoon 2fa: A Static Analysis Approach To Threat Intelligence And Automated Defense, Daniel A. Austin Jr

Cybersecurity Undergraduate Research Showcase

It's gotten much easier to be a cybercriminal. We're seeing a boom in "Phishing-as-a-Service" (PaaS) platforms, which sell advanced phishing attacks as a ready-to-use product. This means almost anyone can now get the tools to launch sophisticated attacks, even if they don't have a lot of technical skill.

This research dives into one of the most prominent threats, the Tycoon 2FA phishing kit. This kit is dangerous because it's designed to bypass Multi-Factor Authentication (MFA) using what is known as an Adversary-in-the-Middle (AiTM) attack.

This paper covers how I built and tested a set of Python-based tools to perform "static …

Reducing Data Requirements In Polymer Science: Deep Neural Networks For Predicting Surface Tension Of Copolymer Compatibilizers, Md Mushfiqul Islam

USF Tampa Graduate Theses and Dissertations

In polymer chemistry, compatibilization involves adding a substance often a block or graft copolymerto stabilize polymer blends that would otherwise not mix well, leading to rough structures and weak me- chanical properties. Compatibilizers improve miscibility and reduce interfacial tension, which is critical for applications such as mixed-waste polymer recycling. Sequence-controlled polymers offer unique potential by combining the tunable chemistry of synthetic polymers with the precise, function-driven design of biological macromolecules, but unlike proteins, they lack large, evolution-shaped datasets to guide discovery. This research develops a deep learning framework to predict the surface tension of sequence-controlled copolymer compatibilizers across varying concentrations. …

Speculative Automated Refactoring Of Imperative Deep Learning Programs To Graph Execution, Raffi T. Khatchadourian Ph.D., Tatiana Castro Vélez, Mehdi Bagherzadeh, Nan Jia, Anita Raja

Publications and Research

Efficiency is essential to support ever-growing datasets, especially for Deep Learning (DL) systems. DL frameworks have traditionally embraced deferred execution-style DL code---supporting symbolic, graph-based Deep Neural Network (DNN) computation. While scalable, such development is error-prone, non-intuitive, and difficult to debug. Consequently, more natural, imperative DL frameworks encouraging eager execution have emerged but at the expense of run-time performance. Though hybrid approaches aim for the "best of both worlds," using them effectively requires subtle considerations. Our key insight is that, while DL programs typically execute sequentially, hybridizing imperative DL code resembles parallelizing sequential code in traditional systems. Inspired by this, we …

Correcting Class Imbalance Through Synthetic Training Data And 3d Modeling For Carabid Pitfall Trap Sampling, Blair Mirka

Geography ETDs

Crowdsourced biodiversity data provide an accessible foundation for large-scale ecological monitoring, but class imbalance limits automated species identification, particularly for rare taxa. This research explores the use of synthetic training data generated from 3D models of carabid beetle museum specimens to improve detection and classification performance for underrepresented species in crowdsourced datasets. High-resolution 3D models were created to simulate variation in lighting, orientation, and background. These synthetic images were incorporated into convolutional neural network training datasets at varying synthetic-to-real ratios to assess their impact on classification accuracy. Models were evaluated using controlled pitfall-trap imagery to examine the influence of scene …

Soccer In-Game Event Classification Using Spatio-Temporal Data, Million Haileyesus

Theses and Dissertations

Classifying soccer ball events, such as pass, shot, ball loss, and ball out, are crucial for advancing game analytics and tactical insights. This thesis investigates the application of machine learning to classify these ball events using player and ball spatio-temporal data, as well as additional features. We implement and compare traditional machine learning algorithms (AdaBoost, Logistic Regression, and Random Forest) with several deep learning approaches, including Feed-Forward Neural Network (FFNN), sequence-to-sequence (seq2seq) recurrent models (Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU)), and Transformer. Our experiments, evaluated on a dataset comprising of three professional soccer matches using accuracy, precision, …