Defense Notices

This research thesis seeks to use machine learning on data from invasive brain-computer interfaces (BCIs) in rhesus macaques to predict their state of movement during center-out tasks. Our research team breaks down movements into discrete states and analyzes the data using Linear Discriminant Analysis (LDA). We find that a simplified model that ignores the biological systems unpinning it can still detect the discrete state changes with a high degree of accuracy. Furthermore, when we account for underlying systems, our model achieved high levels of accuracy at speeds that ought to be imperceptible to the primate brain.

The rapid rise of automated decision-making systems has left a gap in researchers’ understanding of how developers and consumers balance concerns about the privacy and accuracy of such systems against their utility.  With our goal to cover a broad spectrum of concerns from various angles, we initiated two experiments on the perceived benefit and detriment of interacting with automated decision-making systems. We refer to these two experiments as the Patch Wave study and Automated Driving study. This work approaches the study of automated decision making at different perspectives to help address the gap in empirical data on consumer and developer concerns. In our Patch Wave study, we focus on developers’ interactions with automated pull requests that patch widespread vulnerabilities on GitHub. The Automated Driving study explores older adults’ perceptions of data privacy in highly automated vehicles. We find quantitative and qualitative differences in the way that our target populations view automated decision-making systems compared to human decision-making. In this work, we detail our methodology for these studies, experimental results, and recommendations for addressing consumer and developer concerns.

Bankruptcy prediction helps to evaluate the financial condition of a company and it helps not only the policymakers but the investors and all concerned people so they can take all required steps to avoid or to reduce the after-effects of bankruptcy. Bankruptcy prediction will not only help in making the best decision but also provides insight to reduce losses. The major reasons for the business organization’s failure are due to economic conditions such as proper allocation of resources, Input to policymakers, appropriate steps for business managers, identification of sector-wide problems, too much debt, insufficient capital, signal to Investors, etc. These factors can lead to making business unsustainable. The failure rate of businesses has tended to fluctuate with the state of the economy. The area of corporate bankruptcy prediction attains high economic importance, as it affects many stakeholders. The prediction of corporate bankruptcy has been extensively studied in economics, accounting, banking, and decision sciences over the past two decades. Many traditional approaches were suggested based on hypothesis testing and statistical analysis. Therefore, our focus and research are to come up with an approach that can estimate the probability of corporate bankruptcy and by evaluating its occurrence of failure using different machine learning models such as random forest, Random forest, XGboost, logistic method and choosing the one which gives highest accuracy. The dataset used was not well prepared and contained missing values, various data mining and data pre-processing techniques were utilized for data preparation. We use models such asRandom forest, Logistic method, random forest, XGBoost to predict corporate bankruptcy earlier to the occurrence. The accuracy results for accurate predictions of whether an organization will go bankrupt within the next 30, 90, or 180 days, using financial ratios as input features. The XGBoost-based model performs exceptionally well, with 98-99% accuracy.

Having control over nano- and micro-sized objects' position inside a suspension is crucial in many applications such as: sorting and delivery of particles, studying cells and microorganisms, spectroscopy imaging techniques, and building microscopic size lattices and artificial structures. This control can be achieved by judiciously engineering optical forces and light-matter interactions inside colloidal suspensions that result in optical trapping. However, in the current techniques, to confine and transport particles in 3D, the use of high-NA (Numerical Aperture) optics is a must. This in turn leads to several disadvantages such as alignment complications, lower trap stability, and undesirable thermal effects. Hence, here we study novel optical trapping methods such as asymmetric counter-propagating beams where we have engineered the optical forces to overcome the aforementioned limitations. This system is significantly easier to align as it uses much lower NA optics which creates a very flexible manipulating system. This new approach allows the trapping and transportation of different shape objects, sizing from hundreds of nanometers to hundreds of micrometers by exploiting asymmetrical optical fields with higher stability. In addition, this technique also allows for significantly longer particle trapping lengths of up to a few millimeters. As a result, we can apply this method to trapping much larger particles and microorganisms that have never been trapped optically before. Another application that the larger trapping lengths of the proposed system allow for is the creation of 3D lattices of microscopic size particles and other artificial structures, which is one important application of optical trapping.  

This system can be used to create a fully reconfigurable medium by optically controlling the position of selected nano- and micro-sized dielectric and metallic particles to mimic a certain medium. This “table-top” emulation can significantly simplify our studies of wave-propagation phenomena on transmitted signals in the real world. 

Furthermore, an important application of an optical tweezer system is that it can be combined with a variety of spectroscopy and microscopy techniques to extract valuable, time-sensitive information from trapped entities. In this research, I plan to integrate several spectroscopy techniques into the proposed trapping method in order to achieve higher-resolution images, especially for biomaterials such as microorganisms.  

This research project aims to predict the outcome of the Naval Discharge Review Board decision for an applicant based on factors in the application, using Machine Learning techniques. The study explores three popular machine learning algorithms: MLP, Adaboost, and KNN, with KNN providing the best results. The training is verified through hyperparameter optimization and cross fold validation.

Additionally, the study investigates the ability of ChatGPT's API to classify the data that couldn't be classified manually. A total of over 8000 samples were classified by ChatGPT's API, and an MLP model was trained using the same hyperparameters that were found to be optimal for the 3000 size manual sample.The model was then tested on the manual sample. The results show that the model trained on data labeled by ChatGPT performed equivalently, suggesting that ChatGPT's API is a promising tool for labeling in this domain.

RNA molecules play a crucial role in the regulation of gene expression and other cellular processes. Understanding the three-dimensional structure of RNA is essential for predicting its function and interactions with other molecules. One key feature of RNA structure is the presence of base pairs, where nucleotides i.e., adenine(A), guanine(G), cytosine(C), and uracil(U), form hydrogen bonds with each other. The limited availability of high-quality RNA structural data combined with associated atomic coordinate errors in low resolution structures, presents significant challenges for extracting important geometrical characteristics from RNA's complex three-dimensional structure, particularly in terms of base interactions.

In this study, we propose an approach for annotating base-pairing interactions in low-resolution RNA structures using machine learning (ML) based classifiers and leveraging the more precise structural information available in high-resolution homologs to annotate base-pairing interactions in low-resolution structures. We first use DSSR tool to extract annotations of high-resolution RNA structures and extract distances of atoms of interacting base pairs. The distances serve as features, and 12 standard annotations are used as labels for our ML model. We then apply different ML classifiers, including support vector machines, neural networks, and random forests, to predict RNA annotations. We evaluate the performance of these classifiers using a benchmark dataset and report their precision, recall, and F1-score. Low-resolution RNA structures are then annotated based on the sequence-similarity with high-resolution structures and the corresponding predicted annotations.

For future aspects, the presented approach can also help to explore the plausible base pair interactions to identify conserved motifs in low-resolution structures. The detected interactions along with annotations can aid in the study of RNA tertiary structures, which can lead to a better understanding of their functions in the cell.

The Black Literature Network Project is a comprehensive initiative to disseminate literature knowledge to students, academics, and the general public. It encompasses four distinct portals, each featuring content created and curated by scholars in the field. These portals include the Novel Generator Machine, Literary Data Gallery, Multithreaded Literary Briefs, and Remarkable Receptions Podcast Series. My significant contribution to this project was creating a standalone website for the Current Archives and Collections Index that offers an easily searchable index of black-themed collections. Additionally, I was exclusively responsible for the complete development of the novel generator tool. This application provides customized book recommendations based on user preferences. As a part of the History of Black Writing (HBW) Program, I had the opportunity to customize an open-source annotation tool called Hypothesis. This customization allowed for its use on all websites related to the Black Literature Network Project by the end users. The Black Book Interactive Project (BBIP) collaborates with institutions and groups nationwide to promote access to Black-authored texts and digital publishing. Through BBIP, we plan to increase black literature’s visibility in digital humanities research.

With the increased adoption of complex machine learning algorithms across many different fields, powerful computing platforms have become necessary to meet their computational needs. Multicore platforms are a popular choice as they provide greater computing capabilities and can still meet different size, weight, and power (SWaP) constraints. However, contention for shared hardware resources between multiple cores remains a significant challenge that can lead to interference and unpredictable timing behaviors. Furthermore, this contention can be intentionally induced by malicious actors with the specific goals of delaying safety-critical tasks and jeopardizing system safety. This is done by performing Denial-of-Service (DoS) attacks that target shared resources such that the other cores in a system are unable to access them. When done properly, these shared resource DoS attacks can significantly impact performance and threaten system stability. For example, DoS attacks can cause >300X slowdown on the popular Raspberry Pi 3 embedded platform.

Motivated by the inherent risks posed by these DoS attacks, this dissertation presents investigations and evaluations of shared resource contention on multicore platforms, and the impacts it can have on the performance of real-time tasks. We propose various DoS attacks that each target different shared resources in the memory hierarchy with the goal of causing as much slowdown as possible. We show that each attack can inflict significant temporal slowdowns to victim tasks on target platforms by exploiting different hardware and software mechanisms. We then develop and analyze techniques for providing shared resource isolation and temporal performance guarantees for safety-critical tasks running on multicore platforms. In particular, we find that bandwidth throttling mechanisms are effective solutions against most DoS attacks and can protect the performance of real-time victim tasks.

Development and deployment of trusted systems often require definitive identification of devices. A remote entity should have confidence that a device is as it claims to be. An ideal method for fulfulling this need is through the use of secure device identitifiers. A secure device identifier (DevID) is defined as an identifier that is cryptographically bound to a device. A DevID must not be transferable from one device to another as that would allow distinct devices to be identified as the same. Since the Trusted Platform Module (TPM) is a secure Root of Trust for Storage, it provides the necessary protections for storing these identifiers. Consequently, the Trusted Computing Group (TCG) recommends the use of TPM keys for DevIDs. The TCG's specification TPM 2.0 Keys for Device Identity and Attestation describes several methods for remotely proving a key to be resident in a specific device's TPM. These methods are carefully constructed protocols which are intended to be performed by a trusted Certificate Authority (CA) in communication with a certificate-requesting device. DevID certificates produced by an OEM's CA at device manufacturing time may be used to provide definitive evidence to a remote entity that a key belongs to a specific device. Whereas DevID certificates produced by an Owner/Administrator's CA require a chain of certificates in order to verify a chain of trust to an OEM-provided root certificate. This distinction is due to the differences in the respective protocols prescribed by the TCG's specification. We aim to abstractly model these protocols and formally verify that their resulting assurances on TPM-residency do in fact hold. We choose this goal since the TCG themselves do not provide any proofs or clear justifications for how the protocols might provide these assurances. The resulting TPM-command library and execution relation modeled in Coq may easily be expanded upon to become useful in verifying a wide range of properties regarding DevIDs and TPMs.

Remote attestation is a process of establishing trust between various systems on a network. Until now, attestations had to be done on the fly as caching attestations had not yet been solved. With the blockchain providing a monotonic record, this work attempts to enable attestations to be cached. This paves the way for more complex attestation protocols to fit the wide variety of needs of users. We also developed specifications for these records to be cached on the blockchain.