137091 research outputs found
Sort by
Train the Trainer
No full textDiscussion portion of Lost in the Stacks, episode 651. Features discussion about library instruction, and how librarians are instructed to give instruction.Discussion portion of Lost in the Stacks, episode 651. Features discussion about library instruction, and how librarians are instructed to give instruction
Unexpected Applications on IPv6 Ports
No full textAs global IPv6 usage continues to grow, understanding the makeup of services and misconfigurations on the IPv6 address space is crucial to securing the next generation of networks. Prior work on IPv4 has shown that there are many “unexpected” applications running on non-standard ports, but whether this is true in IPv6 remains unexplored. In this thesis, we extend Stanford LZR fingerprinting tool to run IPv6, and integrate it with Georgia Tech’s scanv6 scanner and various address generation techniques to perform internet-wide scans of 55 ports. Between March and April 2025, we discovered over 102 million responsive service endpoints and classified them as “Expected,” “Unexpected,” or “Unknown.” We find that IPv6 contains higher rates of expected-port services compared to IPv4, yet many ports contain mis assignments of services. Our results highlight the effectiveness of current IPv6 address generating techniques and their applications in internet scanning. This work lays the groundwork for IPv6 vulnerability scanning and possible best practices for large-scale internet scanning in the IPv6 era.UndergraduateComputer Scienc
Tuning Viscoelastic and Mechanical Properties in Advanced Thermosets: The Role of Stoichiometry, Surface Modification, and Hybrid Synthesis
No full textThis dissertation presents an integrated materials design strategy for tuning the mechanical, thermal, and viscoelastic properties of advanced thermosets. The research addresses the growing demand for high-performance, reprocessable polymers through three complementary investigations focused on covalent adaptable networks (CANs), its composite systems, and organic-inorganic hybrids development.
The first study investigates Aza-Michael chemistry based CAN, demonstrating that the stoichiometric balance between acrylate and amine functional groups is a critical parameter for controlling stress relaxation and reprocessability. Spectroscopic and mechanical analyses reveal that while systems with balanced stoichiometry maintain mechanical integrity over multiple reprocessing cycles, off-stoichiometry compositions can be engineered for faster network rearrangement at the cost of mechanical robustness.
The second part of the work introduces an interfacial engineering strategy to modulate stress relaxation by incorporating surface-modified silica nanoparticles into the CAN matrix. While amine-functionalized fillers enhanced stress relaxation, the effect was found to be highly dependent on the matrix stoichiometry rather than the alkyl chain length of the surface ligands. These findings highlight the nuanced interplay between filler-matrix interactions and network dynamics, where fillers simultaneously provide chemical pathways to accelerate bond exchange while imposing physical constraints that increase the energy barrier for relaxation.
Finally, the dissertation presents a novel organic–inorganic hybrid epoxy system using aluminum isopropoxide (AIP) as a dual-function agent that acts as both a polymerization initiator and a structural crosslinker, eliminating the need for conventional hardeners. A custom high-pressure curing method was developed to fabricate dense, homogeneous materials. Kinetic analysis confirmed an autocatalytic reaction mechanism, and an optimal AIP concentration was identified to yield superior thermal and mechanical properties.
Together, these studies illustrate a comprehensive design strategy that provides fundamental insights and practical tools for developing the next generation of reprocessable, high-performance thermosets.Ph.D.Chemical and Biomolecular Engineerin
PP7 Virus-like Particles as a Biotechnology Platform
No full textVirus-like particles (VLPs) have found applications in and across many different fields, aided by their stability, capacity for large-scale production, ability for multivalent display of foreign molecules, and inherent immunogenicity. This dissertation explores the versatile utility of VLPs across three distinct areas. The first study investigates the assembly tolerance of engineered dimeric PP7 VLPs by incorporating randomized 15-mer peptide sequences into their progenitor capsid proteins. Using a library-based approach with two iterative rounds of selection, we identified inherent preferences for hydrophilic peptides and for negatively charged amino acids (Asp and Glu) while avoiding bulky and hydrophobic residues like Trp, Phe, Tyr, and Cys. This work led to a considerable expansion of unique assembly-competent VLP sequences, demonstrating the adaptability of PP7-based VLPs to genetic and chemical modifications. The second study examines the structural diversity of VLPs derived from Leviphage PP7, revealing a range of capsid assembly morphologies influenced by small changes in coat protein sequences. This structural plasticity, observed in both native and dimeric forms, suggests that self-assembling structures may naturally exhibit more variability than previously believed, highlighting their capacity to evolve. Lastly, we developed VLP-based vaccines to address emerging variants of SARS-CoV-2. By displaying key peptide sequences from the receptor binding domain (RBD) on VLPs and using them in a heterologous prime-boost regimen with recombinant RBD, we successfully elicited the generation of mutant-biased antibodies in BALB/c mice. This approach identified novel monoclonal antibodies with high affinities towards critical mutations in SARS-CoV-2 variants such as Alpha, Beta, Delta, and Omicron. Overall, these studies collectively demonstrate the broad potential of VLPs for assembly-based applications, structural diversity, and adaptive vaccine development to meet the challenges posed by evolving viral threats.Ph.D.Chemistry and Biochemistr
Crack propagation analysis using pavement image registration and crack vector model for predictive and precision pavement maintenance
No full textPavement maintenance is a crucial aspect of infrastructure management, impacting both the economic competitiveness of a country and the quality of life for its citizens. With over four million miles of pavement in the U.S., and more than $83 billion allocated annually for maintenance, rehabilitation, and reconstruction, there is a pressing need for cost-effective maintenance strategies. This study addresses significant gaps in predictive and precision pavement maintenance by leveraging multi-temporal pavement range images collected by 3D laser technology, aiming to achieve more cost-effective decisions in pavement maintenance and asset management. The study proposes a novel three-stage coarse-to-fine pavement image registration (PIR) methodology, an innovative crack vector model (CVM) based on the crack fundamental element (CFE) concept for flexibly storing and accurately extracting crack properties, and a comprehensive framework for the large-scale implementation of these technologies in a real-world environment. Additionally, it explores crack propagation monitoring and forecasting at both topological and aggregated levels using multi-temporal image registration and fine interval crack properties data. We assess the effectiveness of these methodologies over a 12-year dataset from a 5.8-mile section of US-80 near Savannah, GA. Results indicate that the developed methodologies significantly enhance the accuracy of pavement image registration and crack growth forecasting, supporting the implementation of predictive and precision maintenance practices following the 3R principle—right treatment, right location, right time. This approach ensures the efficient allocation of maintenance resources by precisely determining the highest priority spots for isolated, expensive treatments, such as deep patching, based on not only distress severity but also deterioration rate, to achieve the highest return on investment.Ph.D.Civil Engineerin
Investigating the Alignment of AI Evaluation Processes with Human-Centered Design Principles and National Security Imperatives
No full textThe intersection of national security imperatives and human-centered design (HCD) principles is critical for developing trustworthy artificial intelligence (AI) systems. The adoption of AI technologies in national security has accelerated, yet their alignment with HCD principles remains a significant challenge. Transparency, fairness, and trust in AI systems are necessary to ensure ethical and effective use, especially when these systems impact high-stakes decision-making.
This study aims to investigate how integrating HCD principles can improve the transparency, fairness, and ethical alignment of AI systems within the national security domain. By examining AI systems from both a technical and human-centered perspective, this research seeks to contribute to the development of more reliable and trustworthy AI solutions. Prior studies, such as those by Ozmen Garibay et al. (2023), have emphasized the need for such integration, but gaps remain in how these systems align with specific security and ethical
considerations.
This thesis will explore these gaps by analyzing existing literature, reviewing AI systems currently in use, and applying thematic analysis to evaluate the alignment between HCD
principles and national security requirements.UndergraduateComputer Scienc
Christmas Specials (LITSmas 2025)
No full textDiscussion portion of Lost in the Stacks, episode 667. In the annual "LITS-mas" episode, the Lost in the Stacks hosts discuss the significance of three classic holiday TV specials: Rudolph the Red Nosed Reindeer, A Charlie Brown Christmas, and How the Grinch Stole Christmas.Discussion portion of Lost in the Stacks, episode 667. In the annual "LITS-mas" episode, the Lost in the Stacks hosts discuss the significance of three classic holiday TV specials: Rudolph the Red Nosed Reindeer, A Charlie Brown Christmas, and How the Grinch Stole Christmas
Effect Of Plasma-Cracked Selenium on the Molecular Beam Epitaxy Synthesis of In2Se3 Thin Films
No full textIndium sesquiselenide (In2Se3) is a chalcogenide semiconductor material with broad novel applications in optoelectronics, non-volatile memory, and advanced neuromorphic computing due to its multiple stable phases, most notably, β-In2Se3, γ-In2Se3, and κ-In2Se3. Key issues with synthesizing In2Se3 involve a lack of scalability of pure-phase material and the incorporation of defects due to high-temperature synthesis requirements. Traditional molecular beam epitaxy (MBE) depositions of In2Se2 use thermal evaporation to sublimate Indium and Selenium precursors before deposition, resulting in low phase selectivity, unreacted Selenium clusters, and poor film morphology due to limited surface reactivity.
To address these limitations, this work introduced and assessed the use of a radiofrequency plasma-cracker to break up Selenium into small diatomic or monatomic constituents to generate highly reactive species before deposition. This shift aimed to lower the temperature barriers required to form single-phase β-In2Se3, γ-In2Se3, and κ-In2Se3 while improving their surface morphologies. In this work, a comprehensive comparison using Raman spectroscopy and atomic-force microscopy (AFM) was made between films grown using thermally evaporated Selenium and plasma-cracked Selenium, with Indium consistently produced using thermal evaporation. Further, nucleation and growth between 2-minute and 60-minute depositions were assessed to provide early-stage analysis.
The results of this work show that plasma-cracked selenium showed no improvement or impact on the nucleation characteristics of low-temperature (400°C) films and led to the formation of significant Selenium clustering at longer deposition times. Further, high-temperature films (700°C) deposited using plasma-cracked Selenium led to irregular Selenium cluster formations and Indium-rich mixed phases, likely due to selenium source depletion and plasma instability. Intermediate deposition temperatures (500°C) using plasma-cracked Selenium led to single-phase β-In₂Se₃ growth, pointing towards the merit of this method to modestly lower temperature requirements for In2Se3 formation.UndergraduateMaterials Science and Engineerin
President's Newsletter
No full textAmid so much public debate and critique about the role and value of universities this past year, I was delighted that Forbes chose to write a great story about how Georgia Tech has become a leading outlier in terms of the return on investment for our students and our impact in our community
Using Machine Learning to Fill in Missing Values in Pulsative Data from Diverse Clinical Datasets
No full textThe integration of machine learning techniques with healthcare data has spurred innovative solutions for clinical diagnosis and patient care. This study addresses the critical challenge of missing data in physiological signals, particularly Electrocardiogram (ECG) and Photoplethysmography (PPG), which are vital for diagnosing conditions like atrial fibrillation and monitoring patient health. While existing research predominantly relies on the MIMIC-III dataset, which offers rich but limited patient diversity, this study leverages the MODS dataset from Emory University, providing a more comprehensive representation of patient demographics and clinical conditions. The primary objective is to test the generability of the performance of the BDC Transformer, a machine-learning model for imputing missing values in waveform data, on the MODS dataset. This project underscores the significance of interdisciplinary collaboration in advancing healthcare informatics, with implications for improved patient outcomes and personalized care delivery.UndergraduateComputer Scienc