Treasures @ UT Dallas
Not a member yet
7697 research outputs found
Sort by
Enhancing REST API Access Control Using Multiple Factor Authentication With Refresh Token
No full textRepresentational State Transfer Application Programming Interfaces (RESTful APIs) have emerged as a crucial component in modern web applications, facilitating efficient data exchange between clients and servers and to request data process as well. Securing these APIs is paramount, with a focus on safeguarding endpoints (REST APIs). Despite the impossibility of achieving complete system security, addressing vulnerabilities, particularly in the context of public clients and exchange of access-token and refresh-token, is significant. This thesis explores the utilization of Identity Providers protocol, such as OAuth 2.0 and OpenID Connect, to bolster access control through standardized authentication and authorization methods and guidelines. In the realm of API-driven applications protecting resources and data is not optional. The need for robust security measures has intensified. A notable challenge in RESTful API security lies in authentication and authorization, with refresh-tokens serving as a common means for clients to acquire new access-tokens without user credentials. However, the extended lifespan of refresh-tokens poses a potential security risk if compromised. This research proposes the integration of
multi-factor authentication during the use of refresh-tokens, enhancing account security and mitigating the risk of unauthorized access. This approach offers an additional layer of security without the need to revoke access or refresh-tokens. Given the dynamic nature of access management, staying abreast of the latest developments and best practices is crucial for maintaining application security. This thesis provides a concise overview of key considerations and strategies, leveraging OAuth 2.0 and OpenID Connect, along with a refresh-token-based approach using the Duende.Identity Server as a central Identity Provider. By adopting this approach, developers and organizations can fortify the security of their API-driven applications in the face of an ever-evolving threat landscape. The use of Duende.Identity Server, acknowledged even by Microsoft, ensures the implementation of security measures based on a proven protocol that addresses the concern
In Vivo Transport of Renal Clearable Nanomedicines in Diseased Kidneys
No full textRenal clearable nanomaterials have been considered promising nanomedicines for disease
diagnosis and therapeutics. Understanding the in vivo transport of renal clearable nanomedicines
under various diseased conditions is not only fundamentally important to unravelling
pathophysiology in the nano regime, but also critical to successful translations of nanomedicines
into the clinics. Since kidney disease as a silent killer influence more than 10% people in the
world, understanding the transport mechanisms of renal clearable nanoparticles in the diseased
kidneys will lay down a foundation for developing new disease diagnostics as well as treatments.
This dissertation aims to unravel the in vivo transport of renal clearable nanomaterials in various
kidney disease conditions, which includes seven chapters as follows.
In Chapter 1, an overview is given to the current understandings of nanomedicines and their in
vivo transport in normal and diseased kidney, which are classified into different categories
according to its injury sites and pathologic mechanisms. In Chapter 2, we investigated the in vivo
transport of PEG nanoparticles with different elimination pathways including glomerular
filtration and renal tubular secretion in a well-known kidney disease model, cisplatin-induced
kidney injury. In Chapter 3, we did a head-to-head comparison of the renal tubular secretion of
PEG nanoparticle and small molecule, para-amino hippuric acid (PAH) in cisplatin-induced
renal tubular injuries, and correlation between transport of the exogeneous probes and
biomarkers was also investigated. In Chapter 4, we explored the feasibility of transcutaneous
detection of cisplatin-induced acute kidney injury by combining fluorescent ICG-PEG45 and
IRDye-PEG45 with a commercially available device. In Chapter 5, we studied the in vivo
behaviors of various exogenous probes including the glomerular filtrated FITC-inulin and
Au25SG18 and the renal tubular secreted PAH in cisplatin-injury acute kidney injury. In Chapter
6, we discussed the transport of nanoparticles in other types of kidney disease models including
primary glomerular injuries induced by doxorubicin and tubular obstructions caused by folic
acid. In Chapter 7, an outlook was given based on our current understandings of in vivo transport
and interactions of nanomedicines in kidney diseases
Exploring Machine Learning for Automated Diagnosis in the Presence of Missing and Corrupted Data
No full textThe advances in electronic health records (EHRs) and machine learning (ML) algorithms
have brought a new perspective to biomedical sciences and medical practice. This has
enabled and improved research on automated diagnostics, data-driven disease categorization
and personalized treatments. Researchers and healthcare providers have already welcomed
the recent advancements. However, the transition into practice happens gradually due to
the challenges in the field. Even though patients’ records are being recorded as EHRs in
hospital systems, it is necessary to thoroughly analyze, process, and annotate the data before
employing it for prediction problems. Each chapter of this dissertation highlights the obstacles
that have emerged when implementing learning models on clinical data, particularly in the
application domain of sepsis prediction. The first part of the thesis proposes a strategy to
alleviate the poor prediction performances caused by irregularly-spaced and incompletely
observed databases. The proposed method employs Conditional Generative Adversarial
Networks (GANs), with Long Short-Term Memory (LSTM) networks serving as both the
generator and discriminator, conditioned on class labels. Experimental results show that
while the proposed framework profitably identifies long-term temporal dependencies and
exploits the missing patterns, it also delivers highly notable performance results. The second
part of this dissertation focuses on non-invasive, computationally efficient, and continuous
patient monitoring in intensive care units (ICUs) using single-lead electrocardiogram (ECG)
signals for the early prediction of sepsis. We develop a continuous early sepsis detection
algorithm utilizing two databases; in particular, the Medical Information Mart for Intensive
Care (MIMIC-III) Clinical Database and MIMIC-Waveform Database. We carry out a
systematic approach to selecting ECG segments with superior quality that are recorded in
highly dynamic ICU environments. Moreover, since we are approaching the early sepsis
prediction as a supervised time series classification, we evaluate the model performance
by implementing Temporal Convolutional Networks (TCN). It is discovered that hear rate
variability (HRV) demonstrates considerable decelerations for sepsis patients, and the HRV
characteristics of adults can be a valuable indicator for continuous sepsis monitoring in an
ICU. Finally, this research work adds to the field of early sepsis detection by providing an
annotated continuous waveform database from the MIMIC-Waveform Database, which is
made accessible to the public
Coding for Relay and Wiretap Channels
No full textFor the three-node relay channel, this dissertation investigates the performance of discrete
(coded) modulation in the full-duplex compress-forward (CF) relay channel using multilevel
coding (MLC). Low-density parity check (LDPC) codes are used as the component binary
codes to provide error protection, and the rates assigned to them are numerically analyzed.
For compression at the relay, two methods are utilized: scalar quantization that has the advantage of simplicity, and trellis coded quantization (TCQ) to capture shaping gain. Part of
the contributions of this dissertation is the design of TCQ for end-to-end relay performance,
rather than distortion minimization. A 1 dB gain over prior results for phase shift keying
(PSK) modulation is obtained.
For short-block length relayed communication, this dissertation designs multilevel polar-
coded modulation for the amplify-forward, decode-forward, and compress-forward protocols,
in half-duplex and full-duplex modes. At block length 128 and 256, the presented results are
the first of their kind. At block length 512, improvements of around 2.5dB were obtained
over the state of the art. The design of polar-adjusted convolutional coded modulation
is also presented, in some regimes providing improvements over polar-coded modulation.
Dispersion bounds and error exponent analysis explains and puts in perspective the simulated
performance of the designed coded modulation.
For the wiretap channel, practical design of secrecy codes needs empirical evaluation through
simulations, which has been performed mainly through bit-error rate (BER) simulations.
However, this approach has deficiencies: high BER does not always guarantee zero leakage,
does not give visibility into local weaknesses in the code, and has sensitivity issues due to
the low slope of BER at high error rates. This dissertation proposes a secrecy simulation
metric based on log-likelihood ratios (LLR) and justifies its underlying foundations via its
relationship with equivocation. To give visibility to any local weaknesses, a local version of
this metric is developed, and a principled method is provided for combining local values of
LLR into a single number. For reporting the secrecy performance, the need for specifying the
residual leakage and the confidence level of the simulation is highlighted. A corresponding
confidence interval analysis is provided that is assisted by a density evolution analysis of the
LLR metric. This approach also provides guidance on the number of Monte Carlo samples
needed in simulations, thus helps in the design of secrecy simulations. LDPC and polar
codes for the wiretap channel are simulated and analyzed according to the proposed secrecy
metric. To improve secrecy performance, a serially concatenated coding scheme is proposed
Ensuring Hardware Robustness via Security Verification
No full textSystem-on-Chips (SoCs) play a pivotal role in modern computing systems, integrating multiple Intellectual Property (IP) cores to deliver diverse functionalities. However, this integration presents unique security challenges that can result in vulnerabilities escaping the
verification phase and becoming potential exploits. Moreover, the integration of commercial off-the-shelf (COTS) components into system designs provides cost-effective solutions
but introduces the risk of hidden malicious hardware. On the other hand, asynchronous
events in complex SoC design introduce challenges for security verification. Furthermore,
Network-on-Chip (NoC) architectures introduce new vulnerabilities during message transmission across on-chip networks. Although security properties are introduced to address
these vulnerabilities, generating security properties for SoCs can be a daunting task, typically requiring extensive developer expertise and time. This dissertation extends and explores various approaches to current hardware verification in multiple aspects. First, we
introduce RTL-ConTest, a Register Transfer Level (RTL) security vulnerability detection
algorithm, that extracts critical process flows from a design and executes RTL-level Concolic
testing to generate security test cases for identifying critical exploits. Second, we address
the asynchronous resets by extending the concept of control flow graph (CFG) and extraction of reset-controlled events while avoiding combinatorial explosion. Third, by utilizing
CFG extraction and security properties, we develop a framework for systematic detection
of security violations in NoC designs resulting from vulnerabilities in NoC communication
through formal state exploration. Fourth, We propose an information tracking framework to
identify potential information flow violations in COTS integrated circuits by analyzing their
designs and demonstrating their effectiveness in experimental results. Lastly, we introduce
a language-based machine-learning framework that extracts essential security information
from processor documentation and converts them into security constraints at the RTL level,
enhancing the robustness and efficiency of security property generation
Sound Source Localization for Improving Hearing Aid Studies Using Mobile Platforms
No full textMicrophone array is one of the powerful techniques that enables to apply effective signal
processing algorithms to systems. One of the critical application areas of microphone array
is sound source localization (SSL), which refers to identify the speaker of interest using a
microphone array. SSL can be used as a preprocessing technique to boost up the entire
system efficiency. Recent studies show that smartphones can be an efficient assistive device
for hearing aid devices because of smartphones’ powerful hardware and software components.
Also, Deep Learning (DL) has shown a considerable performance increase in audio signal
processing. DL based SSL using the direction of arrival estimation (DOA) methods for
two and eight microphone array structures and the distance estimation methods using a
single microphone are proposed in this work. The performance of the proposed methods
are evaluated in several realistic noisy conditions, reverberations using real-recorded data.
Another contribution of this work is to present real-time implementations of the DL based
methods on edges devices, i.e., smartphones, tablets
Identification of Urinary Inflammatory Biomarkers for Improved Diagnosis of Urinary Tract Infection in Postmenopausal Women
No full textUrinary tract infection (UTI) is a significant health burden worldwide. UTI has a high rate of
recurrence (rUTI), particularly in postmenopausal women. Diagnosis of UTI and rUTI is
challenging due to poor specificity and a high false-positive rate of current point-of-care (POC)
devices and the 24–72-hour diagnostic window of urine culture – the gold standard of UTI
diagnosis. Most UTIs are treated with antibiotic therapy, which is often prescribed before culturebased diagnosis is complete. The overuse of antibiotics in UTI treatment has contributed to the
rise in antimicrobial resistance among uropathogenic bacteria that reduces the effectiveness of
available antibiotics and complicates UTI and rUTI treatment. To provide more rapid and accurate
UTI diagnosis and reduce the overuse of antibiotics, novel POC diagnostic devices with lower
false-positive rates than the current state of the art devices must be developed. The
pathophysiology of rUTI is a result of an excessive host immune response to uropathogens
colonization. While overactive inflammation plays a key role in pathology and severity of UTI, no
strategies have been developed to leverage inflammatory cytokines for UTI diagnosis or target
inflammation for UTI treatment. The goal of this work was to identify urinary biomarkers for use
as potential diagnostic markers and therapeutic targets for UTI and rUTI in postmenopausal
women. We investigated the role of different inflammatory markers in rUTI in a controlled cohort
of postmenopausal women. We found that urinary levels of inflammatory molecules, PGE2, IL-8,
IL-18, IL-1β, and MCP-1 are significantly elevated in the urine of patients with active rUTI
compared to the patients with no UTI history. Logistic regression analysis revealed that PGE2, IL8, IL-13 combined model was the best performing model in predicting rUTI status. Furthermore,
time-to-relapse analysis indicated that PGE2, but not IL-8 or IL-13, was a strong predictor of UTI
recurrence in postmenopausal women. The risk of rUTI relapse was 3.6 times higher in women
with above-median urinary PGE2 levels than with below-median levels. Taken together, these data
suggest that urinary IL-8, IL-13, and PGE2 may be powerful diagnostic and prognostic biomarkers
for the detection of rUTI in postmenopausal women
Cold Start Active Learning With Submodular Mutual Information for Imbalanced Text Classification
No full textThis study tackles the cold start problem in active learning for imbalanced binary text
classification. Focusing on three datasets (YouTube spam, SMS spam, tweet sentiment) with
class imbalances in their training data, we investigate the efficacy of Submodular Mutual
Information (SMI) methods in the initial active learning stage. These methods aim to balance
class representation using a query set around one percent of the unlabeled data size. We
compare four SMI approaches (two facility location variants, log determinant, graph cut)
with a custom regular expression matching baseline and five established baseline sampling
strategies (Random, BADGE, Entropy, Least confidence, and Margin Sampling) across all
datasets. Our experiments, conducted ten times per dataset reveal that SMI methods,
on average, especially log determinant, outperform both regex matching and traditional
baselines. Further analysis has also been done on the effect of the number of query samples
used on performance. The work highlights the potential of SMI in efficiently addressing the
cold start challenge in imbalanced text classification contexts
Unsupervised Personalization and Deep Uncertainty Modeling for Speech Emotion Recognition
No full textRobust, reliable and generalizable speech emotion recognition (SER) systems have wide application in areas such as healthcare, security and defense. These areas require mission
critical applications where accuracy, test-retest reliability and scalability are of high importance. This dissertation aims to develop SER solutions that can address these requirements.
We focus on two main directions to improve SER performance: personalization of a SER
system to target individuals, and exploring uncertainty modeling to quantify the reliability
of the SER predictions. We formulate the prediction of emotional attributes as a regression
problem and the recognition of primary emotions as a classification problem implemented
using deep neural networks (DNNs).
In our first research direction, we develop personalization approaches for SER models, which
are adapted to improve the performance on target speakers. We study the role of regularization to understand the need for personalization. We focus on the prediction of valence
(negative versus positive), which has often been shown to have lower recognition performance compared to other emotional attributes such as arousal (calm versus active) and
dominance (weak versus strong). Through an exhaustive analysis, we demonstrate that the
prediction of valence needs higher regularization in DNNs than other emotional attributes
such as arousal and dominance. We explore the nature of valence emotional cues conveyed
in speech, finding that they possess stronger speaker dependent traits. Higher regularization
forces the network to learn global patterns that generalize across speakers. This finding suggests that the accuracy of SER models, especially in the valence dimension, can be improved
by developing a personalization strategy that enables the models to capture the speaker
dependent traits from the target subjects. We explore an unsupervised learning approach to
adapt DNN models to target speakers in the test set by searching for speakers in the train
set with similar acoustic patterns as the speaker in the test set. With the data from the
selected speakers in the train set (i.e., the closest speakers to the speakers in the test set),
we propose transfer learning and weighting strategies to adapt the SER system to target
speakers, achieving statistically significant gains in the prediction of valence. This approach
is an effective personalization method for SER problems.
The second direction pursued in this dissertation is the modeling of uncertainty in the SER
predictions. SER is a difficult task when the recordings come from everyday spontaneous
interactions, involving complex emotional behaviors in human communication. Therefore,
it is imperative to develop systems that can provide a confidence score on their predictions,
leading to better calibrated SER systems. We explore approaches for reliability estimation
for SER using Bayesian learning methods, demonstrating the benefits in machine learning
formulation with a reject option. This strategy enables a SER model to abstain from prediction when the confidence of the prediction is low. Therefore, it selectively trades the
coverage on the test set for which the system provide a prediction for better performance.
We propose uncertainty modeling approaches for SER classification and regression tasks using criteria such as empirical risk minimization and methods such as Monte Carlo dropout
(MCD). We analyze the prediction uncertainties as a function of the predicted emotional
attribute scores and the inter-evaluator agreement of the corresponding ground truth annotation. We find that speech segments that are harder for the model to evaluate are harder
for the evaluators to annotate as well. Using probabilistic graphical networks, we demonstrate a novel uncertainty modeling scheme where subjective uncertainties can be learned
using the true annotator label distributions. Likewise, we propose a novel teacher-student
ensemble formulation to perform SER in a scalable and consistent manner that captures
uncertainty in the predictions. Finally, we provide a comprehensive study on the benefits
in uncertainty modeling techniques for SER for the applications of active learning, reject
options and curriculum learning algorithms
Student Haiku Competition 2024
No full textEntries from the 2024 Student Haiku Competition sponsored by the Reference Services Department of The Eugene McDermott Library at The University of Texas at Dallas