Physical Sciences and Mathematics Commons^™

Confronting Barriers To Human-Robot Cooperation: Balancing Efficiency And Risk In Machine Behavior, Tim Whiting

Theses and Dissertations

In strategically rich settings in which machines and people do not fully share the same preferences, machines must learn to cooperate and compromise with people to establish mutually successful relationships. However, designing machines that effectively cooperate with people in these settings is difficult due to a variety of technical and psychological challenges. To better understand these challenges, we conducted a series of user studies in which we investigated human-human, robot-robot, and human-robot cooperation in a simple, yet strategically rich, resource-sharing scenario called the Block Dilemma, a game in which players must balance fairness, efficiency, and risk. While both human-human and …

Symbolic Semantic Memory In Transformer Language Models, Robert Kenneth Morain

Theses and Dissertations

This paper demonstrates how transformer language models can be improved by giving them access to relevant structured data extracted from a knowledge base. The knowledge base preparation process and modifications to transformer models are explained. We evaluate these methods on language modeling and question answering tasks. These results show that even simple additional knowledge augmentation leads to a reduction in validation loss by 73%. These methods also significantly outperform common ways of improving language models such as increasing the model size or adding more data.

Beware Of Ips In Sheep's Clothing: Measurement And Disclosure Of Ip Spoofing Vulnerabilities, Alden Douglas Hilton

Theses and Dissertations

Networks not employing destination-side source address validation (DSAV) expose themselves to a class of pernicious attacks which could be prevented by filtering inbound traffic purporting to originate from within the network. In this work, we survey the pervasiveness of networks vulnerable to infiltration using spoofed addresses internal to the network. We issue recursive Domain Name System (DNS) queries to a large set of known DNS servers world-wide using various spoofed-source addresses. In late 2019, we found that 49% of the autonomous systems we tested lacked DSAV. After a large-scale notification campaign run in late 2020, we repeated our measurements in …

Metalearning By Exploiting Granular Machine Learning Pipeline Metadata, Brandon J. Schoenfeld

Theses and Dissertations

Automatic machine learning (AutoML) systems have been shown to perform better when they use metamodels trained offline. Existing offline metalearning approaches treat ML models as black boxes. However, modern ML models often compose multiple ML algorithms into ML pipelines. We expand previous metalearning work on estimating the performance and ranking of ML models by exploiting the metadata about which ML algorithms are used in a given pipeline. We propose a dynamically assembled neural network with the potential to model arbitrary DAG structures. We compare our proposed metamodel against reasonable baselines that exploit varying amounts of pipeline metadata, including metamodels used …

Trace: A Differentiable Approach To Line-Level Stroke Recovery For Offline Handwritten Text, Taylor Neil Archibald

Theses and Dissertations

Stroke order and velocity are helpful features in the fields of signature verification, handwriting recognition, and handwriting synthesis. Recovering these features from offline handwritten text is a challenging and well-studied problem. We propose a new model called TRACE (Trajectory Recovery by an Adaptively-trained Convolutional Encoder). TRACE is a differentiable approach using a convolutional recurrent neural network (CRNN) to infer temporal stroke information from long lines of offline handwritten text with many characters. TRACE is perhaps the first system to be trained end-to-end on entire lines of text of arbitrary width and does not require the use of dynamic exemplars. Moreover, …

Ansible: Select-To-Edit For Physical Widgets, Benjamin M. Crowder

Theses and Dissertations

Ansible brings select-to-edit functionality to physical widgets. When programming sets of physical widgets, it can be bothersome for a programmer to remember the name of the software object that corresponds to a specific widget. Click-to-edit functionality in GUI programming provides a physical action--moving the mouse to a widget and clicking a button on the mouse--to select a virtual widget. In a similar vein, when programming physical widgets, it is natural to point at a widget and think, "I want to program that one." Ansible allows physical user interface programmers to "click" on a physical widget by making a physical action: …

Light-Field Style Transfer, David Marvin Hart

Theses and Dissertations

For many years, light fields have been a unique way of capturing a scene. By using a particular set of optics, a light field camera is able to, in a single moment, take images of the same scene from multiple perspectives. These perspectives can be used to calculate the scene geometry and allow for effects not possible with standard photographs, such as refocus and the creation of novel views.Neural style transfer is the process of training a neural network to render photographs in the style of a particular painting or piece of art. This is a simple process for a …

Semantically Aligned Sentence-Level Embeddings For Agent Autonomy And Natural Language Understanding, Nancy Ellen Fulda

Theses and Dissertations

Many applications of neural linguistic models rely on their use as pre-trained features for downstream tasks such as dialog modeling, machine translation, and question answering. This work presents an alternate paradigm: Rather than treating linguistic embeddings as input features, we treat them as common sense knowledge repositories that can be queried using simple mathematical operations within the embedding space, without the need for additional training. Because current state-of-the-art embedding models were not optimized for this purpose, this work presents a novel embedding model designed and trained specifically for the purpose of "reasoning in the linguistic domain".Our model jointly represents single …

Cybersecurity Education In Utah High Schools: An Analysis And Strategy For Teacher Adoption, Cariana June Cornel

Theses and Dissertations

The IT Education Specialist for the USBE, Brandon Jacobson, stated:I feel there is a deficiency of and therefore a need to teach Cybersecurity.Cybersecurity is the “activity or process, ability or capability, or state whereby information and communications systems and the information contained therein are protected from and/or defended against damage, unauthorized use or modification, or exploitation” (NICE, 2018). Practicing cybersecurity can increase awareness of cybersecurity issues, such as theft of sensitive information. Current efforts, including but not limited to, cybersecurity camps, competitions, college courses, and conferences, have been created to better prepare cyber citizens nationwide for such cybersecurity occurrences. In …

A Shared-Memory Coupled Architecture To Leverage Big Data Frameworks In Prototyping And In-Situ Analytics For Data Intensive Scientific Workflows, Alexander Michael Lemon

Theses and Dissertations

There is a pressing need for creative new data analysis methods whichcan sift through scientific simulation data and produce meaningfulresults. The types of analyses and the amount of data handled by currentmethods are still quite restricted, and new methods could providescientists with a large productivity boost. New methods could be simpleto develop in big data processing systems such as Apache Spark, which isdesigned to process many input files in parallel while treating themlogically as one large dataset. This distributed model, combined withthe large number of analysis libraries created for the platform, makesSpark ideal for processing simulation output.Unfortunately, the filesystem becomes …

Measuring Influence On Linear Dynamical Networks, Jaekob Chenina

Theses and Dissertations

Influence has been studied across many different domains including sociology, statistics, marketing, network theory, psychology, social media, politics, and web search. In each of these domains, being able to measure and rank various degrees of influence has useful applications. For example, measuring influence in web search allows internet users to discover useful content more quickly. However, many of these algorithms measure influence across networks and graphs that are mathematically static. This project explores influence measurement within the context of linear time invariant (LTI) systems. While dynamical networks do have mathematical models for quantifying influence on a node-to-node basis, to the …

Deep Synthetic Noise Generation For Rgb-D Data Augmentation, Patrick Douglas Hammond

Theses and Dissertations

Considerable effort has been devoted to finding reliable methods of correcting noisy RGB-D images captured with unreliable depth-sensing technologies. Supervised neural networks have been shown to be capable of RGB-D image correction, but require copious amounts of carefully-corrected ground-truth data to train effectively. Data collection is laborious and time-intensive, especially for large datasets, and generation of ground-truth training data tends to be subject to human error. It might be possible to train an effective method on a relatively smaller dataset using synthetically damaged depth-data as input to the network, but this requires some understanding of the latent noise distribution of …

Multi-Human Management Of A Hub-Based Colony: Efficiency And Robustness In The Cooperative Best M-Of-N Task, John Rolfes Grosh

Theses and Dissertations

Swarm robotics is an emerging field that is expected to provide robust solutions to spatially distributed problems. Human operators will often be required to guide a swarm in the fulfillment of a mission. Occasionally, large tasks may require multiple spatial swarms to cooperate in their completion. We hypothesize that when latency, bandwidth, operator dropout, and communication noise are significant factors, human organizations that promote individual initiative perform more effectively and resiliently than hierarchies in the cooperative best-m-of-n task. Simulations automating the behavior of hub-based swarm robotic agents and groups of human operators are used to evaluate this hypothesis. To make …

Deep Learning For Document Image Analysis, Christopher Alan Tensmeyer

Theses and Dissertations

Automatic machine understanding of documents from image inputs enables many applications in modern document workflows, digital archives of historical documents, and general machine intelligence, among others. Together, the techniques for understanding document images comprise the field of Document Image Analysis (DIA). Within DIA, the research community has identified several sub-problems, such as page segmentation and Optical Character Recognition (OCR). As the field has matured, there has been a trend of moving away from heuristic-based methods, designed for particular tasks and domains of documents, and moving towards machine learning methods that learn to solve tasks from examples of input/output pairs. Within …

After Https: Indicating Risk Instead Of Security, Matthew Wayne Holt

Theses and Dissertations

Browser security indicators show warnings when sites load without HTTPS, but more malicious sites are using HTTPS to appear legitimate in browsers and deceive users. We explore a new approach to browser indicators that overcomes several limitations of existing indicators. First, we develop a high-level risk assessment framework to identify risky interactions and evaluate the utility of this approach through a survey. Next, we evaluate potential designs for a new risk indicator to communicate risk rather than security. Finally, we conduct a within-subjects user study to compare the risk indicator to existing security indicators by observing participant behavior and collecting …

Representation And Reconstruction Of Linear, Time-Invariant Networks, Nathan Scott Woodbury

Theses and Dissertations

Network reconstruction is the process of recovering a unique structured representation of some dynamic system using input-output data and some additional knowledge about the structure of the system. Many network reconstruction algorithms have been proposed in recent years, most dealing with the reconstruction of strictly proper networks (i.e., networks that require delays in all dynamics between measured variables). However, no reconstruction technique presently exists capable of recovering both the structure and dynamics of networks where links are proper (delays in dynamics are not required) and not necessarily strictly proper.The ultimate objective of this dissertation is to develop algorithms capable of …

Moderating Influence As A Design Principle For Human-Swarm Interaction, C Chace Ashcraft

Theses and Dissertations

Robot swarms have recently become of interest in both industry and academia for their potential to perform various difficult or dangerous tasks efficiently. As real robot swarms become more of a possibility, many desire swarms to be controlled or directed by a human, which raises questions regarding how that should be done. Part of the challenge of human-swarm interaction is the difficulty of understanding swarm state and how to drive the swarm to produce emergent behaviors. Human input could inhibit desirable swarm behaviors if their input is poor and has sufficient influence over swarm agents, affecting its overall performance. Thus, …

Emergence Of Collective Behaviors In Hub-Based Colonies Using Grammatical Evolution And Behavior Trees, Aadesh Neupane

Theses and Dissertations

Animals such as bees, ants, birds, fish, and others are able to efficiently perform complex coordinated tasks like foraging, nest-selection, flocking and escaping predators without centralized control or coordination. These complex collective behaviors are the result of emergence. Conventionally, mimicking these collective behaviors with robots requires researchers to study actual behaviors, derive mathematical models, and implement these models as algorithms. Since the conventional approach is very time consuming and cumbersome, this thesis uses an emergence-based method for the efficient evolution of collective behaviors. Our method, Grammatical Evolution algorithm for Evolution of Swarm bEhaviors (GEESE), is based on Grammatical Evolution (GE) …

The Security Layer, Mark Thomas O'Neill

Theses and Dissertations

Transport Layer Security (TLS) is a vital component to the security ecosystem and the most popular security protocol used on the Internet today. Despite the strengths of the protocol, numerous vulnerabilities result from its improper use in practice. Some of these vulnerabilities arise from weaknesses in authentication, from the rigidity of the trusted authority system to the complexities of client certificates. Others result from the misuse of TLS by developers, who misuse complicated TLS libraries, improperly validate server certificates, employ outdated cipher suites, or deploy other features insecurely. To make matters worse, system administrators and users are powerless to fix …

Fine-Grained Topic Models Using Anchor Words, Jeffrey A. Lund

Theses and Dissertations

Topic modeling is an effective tool for analyzing the thematic content of large collections of text. However, traditional probabilistic topic modeling is limited to a small number of topics (typically no more than hundreds). We introduce fine-grained topic models, which have large numbers of nuanced and specific topics. We demonstrate that fine-grained topic models enable use cases not currently possible with current topic modeling techniques, including an automatic cross-referencing task in which short passages of text are linked to other topically related passages. We do so by leveraging anchor methods, a recent class of topic model based on non-negative matrix …

Evaluating An Educational Cybersecurity Playable Case Study, Tanner West Johnson

Theses and Dissertations

The realities of cyberattacks have become more and more prevalent in the world today. Due to the growing number of these attacks, the need for highly trained individuals has also increased. Because of a shortage of qualified candidates for these positions, there is an increasing need for cybersecurity education within high schools and universities. In this thesis, I discuss the development and evaluation of Cybermatics, an educational simulation, or playable case study, designed to help students learn and develop skills within the cybersecurity discipline.

This playable case study was designed to allow students to gain an understanding of the field …

Designing Cybersecurity Competitions In The Cloud: A Framework And Feasibility Study, Chandler Ryan Newby

Theses and Dissertations

Cybersecurity is an ever-expanding field. In order to stay current, training, development, and constant learning are necessary. One of these training methods has historically been competitions. Cybersecurity competitions provide a method for competitors to experience firsthand cybersecurity concepts and situations. These experiences can help build interest in, and improve skills in, cybersecurity.

While there are diverse types of cybersecurity competitions, most are run with on-premise hardware, often centralized at a specific location, and are usually limited in scope by available hardware. This research focuses on the possibility of running cybersecurity competitions, specifically CCDC style competitions, in a public cloud environment. …

Improving The Quality Of Neural Machine Translation Using Terminology Injection, Duane K. Dougal

Theses and Dissertations

Most organizations use an increasing number of domain- or organization-specific words and phrases. A translation process, whether human or automated, must also be able to accurately and efficiently use these specific multilingual terminology collections. However, comparatively little has been done to explore the use of vetted terminology as an input to machine translation (MT) for improved results. In fact, no single established process currently exists to integrate terminology into MT as a general practice, and especially no established process for neural machine translation (NMT) exists to ensure that the translation of individual terms is consistent with an approved terminology collection. …

Flow Adaptive Video Object Segmentation, Fanqing Lin

Theses and Dissertations

We tackle the task of semi-supervised video object segmentation, i.e, pixel-level object classification of the images in video sequences using very limited ground truth training data of its corresponding video. Recently introduced online adaptation of convolutional neural networks for video object segmentation (OnAVOS) has achieved good results by pretraining the network, fine-tuning on the first frame and training the network at test time using its approximate prediction as newly obtained ground truth. We propose Flow Adaptive Video Object Segmentation (FAVOS) that refines the generated adaptive ground truth for online updates and utilizes temporal consistency between video frames with the help …

Security Analysis And Recommendations For Coniks As A Pki Solution For Mobile Apps, George Bradley Spendlove

Theses and Dissertations

Secure mobile apps, including end-to-end encrypted messaging apps such as Whats-App and Signal, are increasingly popular today. These apps require trust in a centralized key directory to automatically exchange the public keys used to secure user communication. This trust may be abused by malicious, subpoenaed, or compromised directories. A public key infrastructure (PKI) solution that requires less trust would increase the security of these commonly used apps.CONIKS is a recent PKI proposal that features transparent key directories which publish auditable digests of the public keys they present to queriers. By monitoring its key every time a new digest is published, …

Toward Real-Time Flip Fluid Simulation Through Machine Learning Approximations, Javid Kennon Pack

Theses and Dissertations

Fluids in computer generated imagery can add an impressive amount of realism to a scene, but are particularly time-consuming to simulate. In an attempt to run fluid simulations in real-time, recent efforts have attempted to simulate fluids by using machine learning techniques to approximate the movement of fluids. We explore utilizing machine learning to simulate fluids while also integrating the Fluid-Implicit-Particle (FLIP) simulation method into machine learning fluid simulation approaches.

Usable Security And Privacy For Secure Messaging Applications, Elham Vaziripour

Theses and Dissertations

The threat of government and corporate surveillance around the world, as well as the publicity surrounding major cybersecurity attacks, have increased interest in secure and private end-to-end communications. In response to this demand, numerous secure messaging applications have been developed in recent years. These applications have been welcomed and publically used not just by political activists and journalists but by everyday users as well. Most of these popular secure messaging applications are usable because they hide many of the details of how encryption is provided. The strength of the security properties of these applications relies on the authentication ceremony, wherein …

Netlight: Cloud Baked Indirect Illumination, Nathan Andrew Zabriskie

Theses and Dissertations

Indirect lighting drastically increases the realism of rendered scenes but it has traditionally been very expensive to calculate. This has long precluded its use in real-time rendering applications such as video games which have mere milliseconds to respond to user input and produce a final image. As hardware power continues to increase, however, some recently developed algorithms have started to bring real-time indirect lighting closer to reality. Of specific interest to this paper, cloud-based rendering systems add indirect lighting to real-time scenes by splitting the rendering pipeline between a server and one or more connected clients. However, thus far they …

Certificate Revocation Table: Leveraging Locality Of Reference In Web Requests To Improve Tls Certificate Revocation, Luke Austin Dickinson

Theses and Dissertations

X.509 certificate revocation defends against man-in-the-middle attacks involving a compromised certificate. Certificate revocation strategies face scalability, effectiveness, and deployment challenges as HTTPS adoption rates have soared. We propose Certificate Revocation Table (CRT), a new revocation strategy that is competitive with or exceeds alternative state-of-the-art solutions in effectiveness, efficiency, certificate growth scalability, mass revocation event scalability, revocation timeliness, privacy, and deployment requirements. The CRT periodically checks the revocation status of X.509 certificates recently used by an organization, such as clients on a university's private network. By prechecking the revocation status of each certificate the client is likely to use, the client …

Fully Convolutional Neural Networks For Pixel Classification In Historical Document Images, Seth Andrew Stewart

Theses and Dissertations

We use a Fully Convolutional Neural Network (FCNN) to classify pixels in historical document images, enabling the extraction of high-quality, pixel-precise and semantically consistent layers of masked content. We also analyze a dataset of hand-labeled historical form images of unprecedented detail and complexity. The semantic categories we consider in this new dataset include handwriting, machine-printed text, dotted and solid lines, and stamps. Segmentation of document images into distinct layers allows handwriting, machine print, and other content to be processed and recognized discriminatively, and therefore more intelligently than might be possible with content-unaware methods. We show that an efficient FCNN with …