Student Research at VTNSI

Covert Communications with Large Language Models

Faculty Mentors: Dr. Maice Dutra Da Costa 

Student Researchers: William B, Noah K

Covert communication refers to the stealthy exchange of information to avoid detection by an adversary. Steganography is the practice of hiding information within a non-secret file, message, or object, and it provides methods to achieve covert communication. Modern steganography has developed techniques to conceal data within a digital image or text file. With the growing adoption of Large Language Models (LLMs), there is more content to hide information than ever before. This project uses LLMs to conceal information in plain text. Message encoding is achieved by splitting the vocabulary in sets that are associated with bits or binary sequences. The generation of the plain text then observes the current vocabulary splitting and restricts the word choice according to the hidden message to be encoded. A different vocabulary splitting can be performed at each step of the text generation, using a pseudorandom sequence that can be replicated with knowledge of the initial seed. The decoding process consists of the inspection of the shared text to associate the sequence of words to bits of a hidden message. The results from this project provide a step towards further understanding the use of LLMs for steganography, an area of increasing relevance to information security.

Quantum Computing for Cybersecurity

Faculty Mentors: Dr. Justin Krometis, Dr. James McClure

Student Researchers: Medha D, Kunal K

Quantum computing uses quantum bits (qubits) to perform calculations much more efficiently than classical computers, which use bits that are either a zero or a one. Qubits can represent both simultaneously due to superposition, enabling quantum computers to process many possibilities at once. This technology promises to revolutionize cybersecurity by breaking classical encryption and creating unbreakable codes, given its ability to execute complex calculations reliably with quantum error correction. Quantum-enhanced machine learning can analyze data faster than traditional methods, improving cyber threat detection, fraud detection, and network security. Additionally, quantum computing is being applied in innovative ways, such as developing quantum-based games and reinforcing cybersecurity measures. As quantum computing advances, it can scale to address more complex problems and datasets, enhancing game strategies and cybersecurity. The project also explores error correction in quantum algorithms to minimize errors and noise, essential for practical quantum computing applications. 

Adoption and Use of Generative AI in T&E

Faculty Mentors: Dr. Paul Wach, Dr. Peter Beiling, Dr. Scott Lucero

Student Researchers: Kaushal B, Cameron C, Michael S

Generative AI's utilization and integration for Test & Evaluation purposes is an ongoing discussion within the defense sector. Critical technologies such as LLMs offer transformative promise, however, these technologies also call for new considerations spanning acquisition, security, policy, and technical domains. In our research, we aim to offer actionable insights into the procurement and practical utilization of LLMs for T&E in order to contribute to ongoing strategies in adopting this emerging technology.

Signals of Opportunity for Doppler-based low-Earth orbit PNT

Faculty Mentors: Dr. William (Chris) Headley, Dr. Michael Buehrer

Student Researcher: Megan M

Sponsored by: Virginia Tech National Security Institute 

Doppler-only positioning from low-Earth orbit (LEO) constellations is a promising alternative for a Global Positioning System (GPS) backup. Previous work has developed a position, navigation, and timing (PNT) solution and defined a corresponding Doppler Geometric Dilution of Precision (D-GDOP) model. However, the full solution requires at least eight satellites to be in view. Simulations have shown that having eight consistent satellites in view is unrealistic for the higher elevation masks necessary for LEO satellites. In some cases as few as four satellites are in view, in others as many as thirty. In cases where many satellites are available, choosing a subset of satellites to use in determining position will reduce processing time. However, minimization criteria for D-GDOP have yet to be defined. Unlike GDOP, D-GDOP is also dependent on the velocity and acceleration of the satellites in view. As such, we can no longer assume that a satellite geometry that would minimize GDOP would also minimize D-GDOP. For cases where fewer than eight satellites are in view, we propose a method to reduce the number of required satellites in view by supplementing spatial diversity with time diversity. 

Autonomous Navigation and Payload Deployment

Faculty Mentors: Dr. Kevin Schroeder, Minzhen Du

Student Researchers: Elias B, Thomas H, Nicholas H, Nicholas H, Sreeauditya M, Ella R

Sponsored by: Raytheon (RTX)

Students participated in an annual design competition amongst universities hosted by Raytheon Technologies, building an autonomous air vehicle (UAV) and autonomous ground vehicle (UGV). The UAV must recognize targets and hit them with a water blast, and the UGV must move at controlled speeds in set or random paths. 

Autonomous Multi Sensor Scanning and 3D Model Generation for Ship Configuration Management

Faculty Mentors: Dr. John Gilbert, Dr. Kevin Schroeder

Student Researcher: Nick A

When a naval ship undergoes a change in configuration, be it damage or a new setup on deck, that requires a new engineering analysis, it enters a pipeline of work done by hand over a period of months. This work includes LIDAR scanning and optical imaging, point cloud data cleanup and processing, CAD modeling, and engineering analyses. The goal of this project is to increase the time efficiency of the naval ship scanning to engineering analysis workflow by automating the point cloud processing and CAD modeling steps using state-of-the-art point cloud processing algorithms and a convolutional neural network for object identification.

Ground Testing of CubeSat Compatible Instrumentation to Observe Nitric Oxide 

Faculty Mentor: Dr. Samantha Parry Kenyon, Dr. Leon Harding, Dr. Scott Bailey

Student Researcher: Neha C, Michael P, Ally H, Nathan H

Sponsored by: National Aeronautics and Space Administration (NASA)

In order to better understand how the Earth and Sun interact, CubeSat compatible instrumentation is being developed at Virginia Tech to observe atmospheric nitric oxide. Prior to a CubeSat mission, ground testing must be conducted to mature the instrumentation to technology readiness level six (TRL-6). This testing will replicate some of the conditions the instrumentation will operate under during a space mission. On the optical side, the test bench mimics the collimated ultraviolet star light transmitted by the nitric oxide which the instrument will be observing. Another facet of the test set up is replicating the vacuum and cryogenic environmental conditions of space, this is done through a custom vessel (dewar) that interfaces with a turbopump and cryocooler. This testing will result in a better understanding of how the instrumentation will perform during a space mission. 

CLOUD-D RF: CLOUD-based Distributed Radio Frequency Spectrum Sensing

Faculty Mentors: William “Chris” Headley, Alyse M. Jones, Dr. Maymoonah Toubeh

Student Researchers: Phillip B, Edvin G, Raj K, Joshua L, Caleb M, Zymmorrah M, Akshay P, Lynn P, Pranav P, Ramzy S, River T, Anushka T, Cora W

We are investigating novel approaches to perform collaborative radio frequency spectrum sensing among heterogeneous sensors leveraging AWS cloud services. Spectrum sensing is critically important in both commercial wireless applications (e.g. dynamic spectrum access applications) as well as for military applications (e.g. for jamming/anti-jamming in adversarial environments). While there has been extensive work in the area of collaborative spectrum sensing in recent years, this research problem area to be tackled is unique for the following reasons: 1) assumed use of state-of-the-art deep neural networks at each sensor, 2) assumed heterogeneity of sensors (meaning each sensor are not unique with the same characteristics/performance), and 3) use of the cloud for coordination and collaboration.

Monopulse Direction Finding Antenna Array

Faculty Mentor: Thomas Krauss

Student Researchers: Maria B, Michael H, Gabriel A, Paul B, Albert E, Daniel F, Samuel A, Pranish A, Jordi B, Ivin B, Noah D, Ayia I, Hafsa K, Sean M, Kidus M

Sponsored by: CACI International

In the rapidly advancing field of radio frequency (RF) technology, the Monopulse Direction Finding Antenna Array is a significant stride toward improving the precision of direction finding and tracking of RF signals. This project showcases the design, construction, and testing of a four-element Elevation/Azimuth direction-finding array, specifically tuned for the 915 MHz frequency band. Utilizing a unique assembly of Yagi-Uda antennas mounted on a custom-engineered structure, the system demonstrates remarkable capabilities in pinpointing and following the movements of a transmitter in real time. This endeavor not only highlights the technical challenges of minimizing signal interference and enhancing structural stability but also emphasizes the integration of monopulse comparator networks for efficient signal processing. Through meticulous assembly and calibration of RF circuitry, along with software development for data analysis and system control, the team has successfully demonstrated the antenna array's ability to track dynamic targets. This project lays the groundwork for further advancements in RF direction finding, with potential applications ranging from drone tracking to emergency locator transmissions, exemplifying the practical application of theoretical principles in addressing real-world challenges. 

A Multi-Agent Reinforcement Learning Testbed for Cognitive Radio Applications

Faculty Mentors: Dr. William “Chris” Headley, Alyse Coulon, Dr. Amos Johnson

Student Researchers: Sriniketh V, Daniel R, Raphael R, Michael G

Sponsored by: Office of the Director of National Intelligence (ODNI)

Technological trends show that Radio Frequency Reinforcement Learning (RFRL) will play a prominent role in the wireless communication systems of the future. Applications of RFRL range from military spectrum dominance to commercial products. Before deploying algorithms for these purposes, they must be trained in a simulation environment to ensure adequate performance. For this reason, we previously created the RFRL Gym: a standardized, accessible tool for the development and testing of reinforcement learning (RL) algorithms in the wireless communications space. This environment leveraged the OpenAI Gym framework and featured customizable simulation scenarios within the RF spectrum. However, the RFRL Gym was limited to training a single RL agent per simulation; this is not ideal, as most real-world RF scenarios will contain multiple intelligent agents in cooperative, competitive, or mixed settings, which is a natural consequence of spectrum congestion. Therefore, through integration with Ray RLLib, multi-agent reinforcement learning (MARL) functionality for training and assessment has been added to the RFRL Gym, making it even more of a robust tool for RF spectrum simulation. This paper provides an overview of the updated RFRL Gym environment. In this work, the general framework of the tool is described relative to comparable existing resources, highlighting the significant additions and refactoring we have applied to the Gym. Afterward, results from testing various RF scenarios in the MARL environment and future additions are discussed.

NEEC Logistics Algorithm Development for Quantum Computers Quadratic Assignment Problem

Faculty Mentor: Thomas Krauss

Student Researchers: Nathan D, Julia S

Our team studies the impact of quantum computing on finding solutions to the NP-hard Quadratic Assignment problem. The problem is to assign n facilities to n different locations with the goal of minimizing the cost of the sum of the distances multiplied by the corresponding flows. Previous research has focused on classical solutions. We describe our quantum approach using neutral atom methods and our work thus far. 

RFRL Gym: A Reinforcement Learning Testbed for Wireless Communications

Faculty Mentors: Alyse Jones, Dr. Amos Johnson, Dr. William “Chris” Headley

Student Researchers: James B, Kai B, Jadez D, Dylan G, Michael G, Raphael R, Sriniketh V

Radio Frequency Reinforcement Learning (RFRL) is anticipated to be a widely applicable technology in the next generation of wireless communication systems, particularly 6G and next-gen military communications. Given this, our research is focused on developing a tool to promote the development of RFRL techniques that leverage spectrum sensing. In particular, the tool was designed to address two cognitive radio applications, specifically dynamic spectrum access and jamming. In order to train and test reinforcement learning (RL) algorithms for these applications, a simulation environment is necessary to simulate the conditions that an agent will encounter within the Radio Frequency (RF) spectrum. In this project, such an environment has been developed, herein referred to as the RFRL Gym. Through the RFRL Gym, users can design their own scenarios to model what an RL agent may encounter within the RF spectrum as well as experiment with different spectrum sensing techniques. As such, we have open-sourced this codebase to enable other researchers to utilize the RFRL Gym to test their own scenarios and RL algorithms, ultimately leading to the advancement of RL research in the wireless communications domain. 

Proposing a Federal Democracy in Burma

Faculty Mentor: Robert Hodges

Student Researchers: Joseph B, Jackson C, Josephine E, Amanda H, Haleigh H, Katherine L, Kyle S, Helena S, Lauren S, Vicky S

Sponsored by: Hume Research Fellowship & Department of State (US Embassy Rangoon Diplomatic Support Unit)

On February 1, 2021, a military coup ended the democratic civilian rule of Myanmar and instituted military rule under the leadership of General Min Aung Hlaing. Since the coup, the nation has been in a civil war with the military junta against the various Myanmar ethnic groups. While the ethnic groups have regained control of much of the nation, the military is instituting conscription, and the Bamar ethnic group are fleeing to avoid such. These events display an opportunity for the rise of a new democratic federal democracy in Myanmar. A federal democracy presents the strongest opportunity for the people of Myanmar to avoid human rights violations, such as the ones against the Rohingya. This proposal needs strong assistance in its early stages, as a nation with lesser capabilities would allow the military junta to regain control as an authoritarian regime. The largest external threat to such is China, so firm, global support is necessary to maintain stability in the nation and region as the federal democracy roots itself firmly for continued, long-term growth. 

Countering Extremism in the Sahel

Faculty Mentor: Robert Hodges

Student Researchers: Cameron P, Liz M, Paris A, Jacky B, Frank B, Lily D, Will E, Amanda G, Maggie G, Jess G, Haleigh H, Aidan M, John N, Allie P, Asher R, Lane R, Bianca S, Ethan W, William Y

Sponsored by: Department of State 

The current crisis in the Sahel, specifically in Mali, Burkina Faso, and Niger, is a mix of multiple issues; tensions between various ethnic groups in the Sahel, decreased economic opportunity, increased terrorist recruiting, and an influx of terrorist groups armed from the conflict in Libya but largely pushed out of the Maghreb (Mauritania, Morocco, Algeria, Libya, Tunisia). In the past three years multiple Sahelian countries have undergone coups ostensibly as a response to a failure by civilian governments to defeat ongoing terrorist insurgencies, further constraining U.S. policy options. This is an examination and exploration of what U.S. policy options outside of pure security solutions remain, and what local solutions are likely to result in positive progress, as well as looking at what factors led to the significant reduction in terrorist activity in the Maghreb, and to what extent those factors are replicable in the Sahel. 

Re-Building the Ukrainian Defense Industrial Base

Faculty Mentor: Robert Hodges

Student Researchers: Patrick L, Elena R, Francine B, Brandon C, Stephen D, Kendall F, ByungHoon H, Megan I, Abby J, Logan K, Daniel M, Ashley M, Emil M, Jake O, Christian R, Jake S, Dillon S, Ezra S, Safieh T, Kaleb T, Cameron U, Dori V, Anthony V 

Sponsored by: Department of State Bureau of Political-Military Affairs

The Department of State Diplomacy Lab Project, "Re-Building the Ukrainian Defense Industrial Base," researched and analyzed the most suitable defense industrial base (DIB) for Ukrainian development and countering future Russian aggression and military operations. Using different forms of intelligence collection, the students determined that the future of the Ukrainian DIB must be placed in close proximity to NATO states in the Western region of Ukraine. This is for three reasons. First, the distance to the frontlines protects the DIB against most conventional attacks, such as artillery and ground assault. Second, the strategic location of the DIB along the Western Ukrainian Oblasts with NATO provides an additional layer of security as targeting these positions increases the risks of NATO escalation. Third, this region of Ukraine would allow Western companies to establish partnerships with the Ukrainian Defense Industry and allow ease of access from NATO into Ukraine. The most prominent threats to this DIB format are airstrikes and unconventional operations, such as cyber-attacks. To counter this, the DIB requires a multi-layered defense system, including air and ground defenses and security procedures. Additional recommendations relating to covert operations, international engagement, and intelligence collection are provided.