Trident Scholar Abstracts 2020

Kaden C. Dohm
Midshipman First Class
United States Navy

Optimizing the Structural and Deflagration Properties of Aluminum-Rich Paraffin-Based Hybrid Rocket Motors

Hybrid rocket motors offer an alternative to conventional solid motors and liquid engines. One of the most severe limiting factors of hybrid rocketry is the low regression rate as the fuel does not burn nearly as quickly as in a solid rocket. However, the discovery of liquefying fuels such as paraffin has greatly improved the regression rate of hybrid rocket fuels. Such fuels operate by liquefying to a greater extent than conventional fuels, whereby the low viscosity allows droplets of the fuel to become entrained in the oxidizer flow which increases the speed of the combustion process. Paraffin has a low compression strength and fractures easily. Recent studies show that the incorporation of polyethylene and aluminum powder can maintain and even increase the high regression rate of paraffin rocket motors, and increase the structural strength to allow these motors to survive the flight loads they experience during launch. This project seeks to further advance the benefits of the addition of aluminum powder as previous research suggests that the addition of aluminum powder at a concentration greater than 25% by mass could greatly increase the regression rate. As the concentration of aluminum powder increases more energy is released from the combustion of the fuel which increases the temperature of the combustion resulting in a faster regression rate, but the fuel also becomes more viscous which prevents the entrainment of fuel droplets which serves to limit the regression rate. Due to the COVID-19 pandemic, test fires were unable to be completed to measure the regression rate of the fuel samples with a higher concentration of aluminum powder. However, data from compressive yield testing of each of the fuels suggests that the optimal fuel composition contains 35% aluminum powder by mass.

---

Android mobile devices use a unique method of authentication in the form of a single-stroke graphical pattern on a 3x3 grid that a user is required to create and recall. In this research project, we are going to explore improved iterations of this Android Pattern Lock in the pursuit of guiding users towards creating more secure patterns. Within the past five years, Mobile Authentication methods have continually progressed towards creating a more secure means to safeguard a mobile device. Such methods now include biometric identification, system assisted password guidance via blacklists, and longer minimum passcode lengths. While many methods have progressed, the standard authentication interface for Android devices remains similar in comparison to its initial model. In this work, we sought to explore the effects of changing the existing Android pattern lock interface to an interface we deemed the Double Pattern.

We examined the methodologies by which users chose their Double Patterns using our new interface, specifically metrics related to the complexity of the patterns created, pattern frequency within each treatment population, usability aspects of the interface itself, security strength of our interface, and perceived security strength related to existing authentication methods. Ultimately, we found that our Double Pattern had a significant increase in security related to lower partial guessing entropy and lower susceptibility to simulated guessing attacks, due to the low occurrence rate of each Double Pattern. Equally important, participants perceived the Double Pattern as a more secure interface than the original interface, specifically within our users who previously utilized Android unlock patterns. We are confident based on these results that the Double Pattern could be feasibly implemented as a progression of the original Android unlock pattern interface.

---

We investigate a novel approach to Proofs of Retrievability (PoR), protocols that allow a client to audit the cloud server storing its data remotely. These protocols allow a means of efficiently ensuring that all of the data the client believes to be stored in the cloud is still able to be retrieved by the cloud server, instead of relying on trust alone in the current model. Past PoR approaches have worked toward computational optimization for the audit; however, this requires a large amount of overhead persistent storage (up to 10x the actual database size). Our new approach instead trades higher computation for significantly decreased persistent storage. As all major cloud providers charge markedly more for storage than for computation, our new protocol offers practical efficiency. Our approach rests on treating the data as a square matrix, comparing randomized linear algebra identity tests over the matrix at the time of last check and at the current time. Honest retrieval of data, enforced through a Merkle hash tree requiring negligible extra persistent storage, and dynamic updates are supported in our approach. While audit computation now scales linearly, the required persistent storage is only 1.068x the size of the data. We demonstrate its efficiency in practice with a deployment on Google Cloud Compute Engine with test case data size of 1TB. Our approach costs $42.72 per month for storage, and an audit costs $0.23 taking 16 minutes. Previous state of the art requiring 6x storage of the data size costs $240 per month. This is a 82% cost savings from storage while hosting the data in the cloud. We parallelized the computation of the audit across multiple virtual machines using MPI in order to increase the I/O-bound run time performance, which resulted in a near-linear speed up. We are investigating further optimizations on client-side storage and communication costs, as well as how to deploy our approach over an entire block device.

FACULTY ADVISOR
Associate Professor Daniel Roche
Computer Science Department

---

Christian E. Hoffman
Midshipman First Class
United States Navy

---

As a beam propagates, it is subject to fluctuations in the refractive index of air. These effects can be modeled as optical turbulence. Optical turbulence limits the effectiveness of laser-based weapons and communication systems employed by the United States Navy. Models developed to predict optical turbulence through the structure constant C _n ² are sensitive to absolute air temperature. Existing models have, however, failed to accurately predict the rapid beam attenuation and corresponding high values of C _n ² observed in maritime and near-maritime environments. In response, data-driven machine learning models were developed to predict the refractive index structure parameter C _n ², and to explore the importance of various environmental factors on its prediction.

The current study uses 15 months of C _n ² field measurements collected along an 890 m scintillometer link over the Severn River at the United States Naval Academy. Measures of optical turbulence are complemented by corresponding measurements of 12 environmental parameters. Fully data-driven models were trained, developed, and tested to enhance C _n ² prediction accuracy in the near-maritime environment. Analysis of these models resulted in better understanding of the relative importance of each environmental parameter in accurately predicting C _n ². To our knowledge, this is the first application of purely data-driven machine learning models for predicting C _n ² in the near-maritime environment.

Both parametric and non-parametric models were investigated. A general linear model, regression tree, random forest, and boosted regression tree model were trained on field observations and tuned to improve their predictive power. Compared to linear models, regression trees, random forests, and boosted models demonstrated higher C _n ² prediction accuracy. The random forest and boosted model were analyzed to determine the relative importance of each measured environmental parameter in predicting C _n ². The absolute air temperature, air-water temperature difference, temporal hour weight, and seasonality were significant in making accurate predictions of C _n ² in the near-maritime environment.

FACULTY ADVISORS
Associate Professor John Burkhardt
Mechanical Engineering Department

Associate Professor Cody Brownell
Mechanical Engineering Department

---

Jamie W. Lee
Midshipman First Class
United States Navy

Oblivious k-Nearest Neighbors for Secure Map Applications

Cloud storage enables users to access and store large amounts of data on servers anytime and anywhere at little or no cost. Map applications are specific examples of cloud storage servers that allow users to query for nearby points of interest. Despite the many benefits offered by map applications, users are susceptible to data leakage through their access patterns, which is a significant security risk for these applications since the user’s location and other sensitive data can be leaked.

In order to mitigate access pattern leakage and implement security in map applications, we have developed a novel remotely-stored network data structure, the ORAM-backed Hilbert B-tree. The novel data structure combines existing features such as the B-tree data structure, k-Nearest Neighbor (k-NN) search algorithms, Hilbert Curves, and Oblivious Random Access Memory (ORAM) algorithms, but ultimately allows users to make oblivious queries in map applications, a function that has not yet been conceived for such applications.

The implementation process involves several steps. First, the methods of the B-tree data structure are modified to enable k-NN queries and return a range of data closest to the input. Next, orthogonal Hilbert space-filling curves are implemented to allow the B-tree to handle two-dimensional (2D) data. Further obfuscation of the data is facilitated through ORAM algorithms, which hide the users’ access patterns by encrypting and shuffling logical locations of data in memory. Finally, the data structure is modified to operate as a cloud server with the use of Google Cloud Services. The final Network ORAM-backed Hilbert B-tree allows the user to conduct k-NN range searches without leaking metadata and for the cloud to store and retrieve data obliviously. This provides a significant improvement in security for map applications without compromising performance, preventing sensitive information such as the user’s physical location from being compromised.

FACULTY ADVISORS
Associate Professor Daniel Roche
Computer Science Department

Assistant Professor Travis Mayberry
Computer Science Department

Associate Professor Adam Aviv
George Washington University

---

Even with the dramatic advances in computational power seen in the last decades, Computational Fluid Dynamics (CFD) models are as yet unable to predict transition, separation, and wake development for fluid flow over three-dimensional bodies to the desired level of accuracy in an acceptable amount of time. Without the ability to predict forces and moments experienced by the body, critical parameters such as drag and loads on control surfaces for air- and water-borne vehicles cannot be predicted. The prolate spheroid has long been a popular body upon which to verify CFD models because of its simple geometry and three-dimensional flow field. Advances in computational speed and experimental capabilities have prompted a renewed interest in related research.

An experiment was conducted in the large towing tank facility of the U.S. Naval Academy, using a 6:1 prolate spheroid, measuring 54 in. (1.4 m) in length and 9 in. (0.23 m) in diameter. The spheroid model was inclined by 15° relative to the undisturbed free surface, and towed at speeds yielding length-based Reynolds numbers from 0.5-4.2×10 ⁶. The results from the 0.5×10 ⁶ case are presented in the present discussion. A stationary stereo particle image velocimetry (SPIV) system was designed for the experiment and used to provide two-dimensional velocity maps in two spatial-dimensions (2C2D). These time histories show the trajectory of the wake as it leaves the tail of the model, the expansion of the wake width, the size, strength, and position of the primary vortical structures shed into the wake. These results will inform follow-on studies focused on measuring turbulent quantities in the far wake.

FACULTY ADVISOR
LCDR Ethan Lust, USN
Mechanical Engineering Department

---

Clayton S. Pelzer
Midshipman First Class
United States Navy

The Effects of Acceleration on Film Cooling in Gas Turbine Engines

Gas turbine engine blades contain film cooling holes that direct cool air over the surface of the blades, protecting them from high temperatures up to 2000 °C inside the turbines. There are many factors that affect film cooling, but one of the most important is the acceleration of flow over the surface of the blade. Its impacts on the film cooling process are not fully understood. Previous experiments have compared film cooling both with and without accelerating flow, but no study has reported on heat transfer and flow with different levels of acceleration. In order to understand the direct effects of acceleration, a test section was constructed to model the interactions of flow over a gas turbine blade by matching dimensionless parameters. The test section has lower temperatures and larger dimensions than an actual turbine blade, physically allowing for direct observation. The values determined through this work should give blade designers a more complete understanding of the heat transfer and flow associated with film cooling of a turbine blade. In the end, this work, combined with others’ work investigating how different factors influence film cooling, should allow engines to be run hotter and more efficiently. This would have far-reaching effects, impacting gas-turbine engines that are used in many different applications.

FACULTY ADVISORS
Professor Ralph Volino
Mechanical Engineering Department

Assistant Professor Ronald Warzoha
Mechanical Engineering Department

---

Aidan J. Sabety-Mass
Midshipman First Class
United States Navy

Decision Problems in Computational Group Theory

In this report we discuss a proof following the ideas of Gabor Elek and Endre Szab that Kaplansky’s conjecture is satisfied for group algebras over finite groups and arbitrary fields; as well as, add detail to a proof that Gottschalk’s conjecture implies Kaplansky’s conjecture over fields with positive characteristics. We further present two algorithms: one is a unique algorithm that, if given an infinite amount of time, can in principle iteratively check Kaplansky’s conjecture for all finite groups. The other is a unique implementation of a paper by Dykema, Heister and Juschenko that can check infinite groups from the class of Universal Left Invertible Element (ULIE) groups.

FACULTY ADVISOR
Associate Professor Kostya Medynets
Mathematics Department