Funded programs 2025

1. Title: Psychological game Theory-Based Optimization Approaches and Virtualization Tools for Networked Computing Systems
PI: Manish Bansal
Lead Institution: Virginia Tech
Co-PIs & Institution: Dr. Alec Smith, Dr. Mark Gardner
Funding Program: FY25 Research Program

Summary: Networked computing systems have found a multitude of applications across various domains, revolutionizing the way we communicate, collaborate, and interact with the world. Despite these advantages, the limited capabilities of existing decision tools for networked computing systems to conduct vulnerability analysis against cyberattacks of a stochastic nature, is hindering the reliance on this technology. With the rapid rise in cyberattacks, it is critical to develop data driven control and decision support tools (algorithms/mathematical/simulation models) for system operators, especially tools that utilize relevant historical data and belief-dependent psychological payoffs of attackers and operators for modeling, and virtualization tools for validation. This project will integrate mathematical optimization techniques with game theory and behavioral economics to address cyber-security and cyber-defense for networked computing systems. Specifically, the project will introduce data driven and computationally efficient decision tools that allow limited availability of relevant historical data to infer patterns, and thereby assess the probability of success and the potential impact of cyberattacks. Novel game theory-based algebraic/optimization models for vulnerability analysis of networked computing systems will be developed. These models will be extended to include psychological payoff terms that capture emotions such as anger, frustration, or guilt. The predictive validity of these payoff terms will also be evaluated. This research will offer a unique advantage of analyzing the impact of potential cyberattacks under a wide range of circumstances. The proposed models and algorithms will be evaluated by emulating the networked computing system using virtualization technology, staged Mirai virus attacks, and real network activity data. 

2. Title: Quantum Information Science and Engineering Activities at Virginia Tech
PI: Ed Barnes
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Research Program

Summary: Virginia Tech's Center for Quantum Information Science and Engineering (VTQ) plays a pivotal role in advancing QISE research through key activities including the semi-annual VTQ Symposium and weekly seminar series. The symposium features faculty and senior postdocs presenting research, fostering interdisciplinary collaborations, and building community within VTQ. The seminar series, organized by an interdisciplinary committee, hosts impactful speakers to enhance QISE visibility and reputation. Additionally, VTQ conducts outreach, including a high-school-level summer school on quantum computing. Budget allocations include $12,000 for seminars, $6,000 for symposiums, and $2,000 for summer school support.

3. Title: Bridging the Gap Between User Interface Security and CI?CD workflows
PI: Chris Brown
Lead Institution: Virginia Tech
Co-PIs & Institution:  None
Funding Program: FY25 Research Program

Summary: Software, a set of instructions to execute tasks and manipulate data on devices, impacts nearly every facet of modern life. Humans interact with software-based applications through user interfaces (UIs) which handle the input and output for programs—such as web applications hosted online and accessed through internet browsers. To enhance the delivery of software to users, continuous integration and continuous deployment (CI/CD) techniques have been introduced to automate tasks for more efficient testing and sharing of software. However, as the complexity and demand for technology increases, user interfaces grow increasingly difficult to design and secure while web-based attacks become more sophisticated. Research findings have posited a wide variety of automated tools to enhance the security of software, such as dynamic application security testing (DAST) tools which can assess the security of web applications by simulating attacks on user interfaces. However, these tools are often avoided in practice—with open source software developers facingdifficulties integrating UI test cases into CI/CD workflows and often relying on more ad hoc and manual approaches to test the security of their software.To this end, the proposed work aims to bridge the gap between UI testing and CI/CD pipelines in the context of securing web applications. We accomplish this by designing, implementing, and evaluating a novel tool that leverages large language models (LLMS) that synthesize DAST tool results to support finding and repairing UI-based security vulnerabilities. This work provides implications and motivates future research to further secure user interfaces and safeguard user experiences with software. 

4. Title: Moving Target Defense for Continuous Operation in Mission-Critical Cyber-Physical Systems
PI: Thidapat (Tam) Chantem
Lead Institution: Virginia Tech
Co-PIs & Institution: Mohamed Azab (VT)
Funding Program: FY25 Research Program

Summary: In this project, we aim to explore the application of Moving Target Defense (MTD) strategies within Real-Time Cyber-Physical Systems (RT-CPS) to enhance their security. MTD involves implementing dynamic and unpredictable changes to the system architecture, making it more challenging for attackers to exploit vulnerabilities. This proactive approach aims to increase the complexity and effort required for an attacker to locate vulnerabilities, thereby consuming significant amounts of their resources and time. The project’s primary objective is to integrate MTD effectively without compromising the operational efficiency of RT-CPS, which often operate under stringent time constraints and resource limitations. We plan to develop a framework that carefully balances the security enhancements provided by MTD with the critical operational demands of real-time systems. This framework will be evaluated using two key CPS scenarios: connected vehicles and smart grids. Both contexts are highly relevant and sensitive to delays and disruptions, making them ideal candidates to test the feasibility and impact of MTD. Through this investigation, we seek to achieve a robust security posture that significantly extends beyond conventional defense mechanisms, adapting to emerging threats while maintaining system integrity and performance.

5. Title: IrisSwap: A public research challenge on digital iris spoofing attacks
PI: Brendan David-John
Lead Institution: Virginia Tech
Co-PIs & Institution: Chris Thomas (VT)
Funding Program: FY25 Research Program

Summary: Video-based eye trackers capture the iris biometric and enable authentication to secure user identity. Iris authentication through eye trackers is becoming standard within virtual and augmented reality headsets. However, iris biometric authentication is susceptible to spoofing another user’s identity through physical or digital manipulation. The current standard to identify physical iris spoofing attacks on eye-tracking sensors uses gaze-based liveness detection. Liveness detection classifies gaze data as real or fake, which is sufficient to detect physical presentation attacks. However, preliminary work has exposed a vulnerability in current liveness detection models to digitally manipulated spoofing attacks. The proposed objectives of this project are to explore a broader set of digital manipulation attacks to compile a benchmarking dataset for the field of iris spoofing, and use this dataset to facilitate a globally visible research challenge and workshop to establish a new state-of-the-art defense mechanism. The results of this challenge dataset and the research deliverables in this proposal will inform future considerations for iris spoofing vulnerabilities and is expected to produce IP for the system developed by the PIs, as well as provide a roadmap for future large external research roposals on securing iris biometrics.

6. Title: Fundamentals of Geotagging Images and Videos Using Non-Unique Landmarks
PI: Harpreet S. Dhillon
Lead Institution: Virginia Tech
Co-PIs & Institution: Michael Buehrer (VT)
Funding Program: FY25 Research Program

Summary: The problem of geolocating videos and images from unknown locations is of fundamental interest in numerous cybersecurity applications. Since large databases of geotagged aerial/satellite images are available, this problem reduces to establishing a mapping between a ground viewpoint and its corresponding geotagged aerial viewpoint. This is a highly challenging problem since different parts of the same object might be occluded in the two viewpoints, making it appear very different in the ground and aerial images. Despite recent advances in deep learning, a deployable solution to this problem has eluded us. The overarching goal of this research effort is to develop a systematic solution to this problem based on the observation that only a small part of any image is typically useful for geolocation. In the absence of any recognizable landmarks, the task reduces to using the relative placement of a combination of distinct types of landmarks, where landmarks of the same type may be indistinguishable. If this relative placement of several landmarks is unique enough on a map, it immediately reduces the number of candidate solutions significantly. Modeling these landmarks as an appropriately chosen marked point process, we study the uniqueness of various combinations of different landmarks through a metric termed localizability, which is the probability that the exact pattern of landmarks observed in the image of interest will appear elsewhere in the dataset or map. These initial results will feed into an external grant application where the point process representations will also inform novel attention mechanisms for cross-domain learning.

7. Title: Addressing Software Defects and Security Vulnerabilities in Smart Home Automations
PI: Xinghua Gao
Lead Institution: Virginia Tech
Co-PIs & Institution:  Na Meng, Daphne Yao
Funding Program: FY25 Research Program

Summary: Smart homes are becoming prevalent. People tend to transform residences into smart homes by customizing off-the-shelf smart home platforms, instead of creating IoT systems from scratch. Among the alternatives, Home Assistant (HA) is the most popular platform and has about one million users around the world, mainly because it is free and open-sourced. HA allows end-users (i.e., home residents) to smartify homes by (S1) integrating selected devices into the system and (S2) creating YAML files to control those devices. Unfortunately, due to the diversity of devices and complexity of automatic configuration, many users have difficulty correctly creating YAML files (i.e., fail in S2). Consequently, their smart homes do not work as expected, causing frustration in users and jeopardizing home safety as well as security. Existing tools provide limited support in helping people develop home automation. This project will create HAC (short for Home Automation Check)—a tool framework to help people address software bugs and security vulnerabilities in YAML files.

Specifically, we will (i) create a debugger to detect and fix bugs in YAML-based automation configuration via static analysis, to address coding issues that prevent automation from execution, (ii) create a simulation-based automatic test generator, to reveal vulnerabilities of safety and security via automatic test generation and simulation, and (iii) hold a workshop on smart home security to invite researchers, share ideas, and promote collaboration. If successful, our research will help HA end-users to easily customize HA to their own smart homes for better life quality, home security, and energy efficiency.

8. Title: Enhancing the Privacy of Users with Disabilities Through Transparent Web Advertisement
PI: Muhammad Ali Gulzar
Lead Institution: Virginia Tech
Co-PIs & Institution: Eugenia Rho (VT)
Funding Program: FY25 Research Program

Summary: Web advertising (ads) fuels free content and services but often comes at the cost of privacy violations. This is especially concerning for users with visual impairment and cognitive disabilities who are deceived into interacting with ads as they struggle to interpret the ads’ true intent. Such involuntary interaction with deceptive ads leaks users' browsing activity to third-party services, which is a clear breach of privacy. This proposal aims to achieve equal access to privacy-critical information necessary for users with disabilities to distinguish ads from benign web content, understand the true content of ads, and the consequences of interacting with ads on the web. To this end, we will first conduct a large-scale mixed-method investigation to assess the ADA compliance of web ads across 100K sample websites. More importantly, we will measure the prevalence of misleading alternate text of accessible ads, which is even more deceiving and harmful. We will design a technique that automatically transforms non-ADA-compliant ads into accessible and transparent ads, disclosing their privacy-invasive operations preemptively in plain language. Utilizing the team’s complementary expertise in tracking Javascript localization and computational linguistics, we propose tracing the flow of sensitive personal data from ads to malicious actors and incorporating its comprehensible description in the ads' alternate text. There are 84 million internet users in the US alone who have some form of disability. This research will be transformative in equipping these internet users with tools to perform safe and informed interactions on the web.

9. Title: Secure Component Monitoring in Open FutureG Networks
PI: Wenjing Lou
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Research Program

Summary: The project aims to ensure operational security and build observability into network components in the 5G and beyond (i.e., FutureG) cellular networks using remote attestation. The FutureG network serves as the information backbone for millions of connected devices, enabling high bandwidth, low latency, greater coverage, and reliable connectivity. The implementation of Open Radio Access Network (ORAN) opens up the interfaces between network components and enables the operators to mix and match components from different vendors. However, this also increases the risk of disruption or malicious attacks that could lead to severe consequences. To address this, full visibility into the network’s operational state is crucial. Remote attestation is a recent development in trusted computing that allows for cryptographic verification of the system’s expected functionality. The PI and her research team have extensive research experience in this field. In this project, we propose a network attestation tool with two key components: a network evidence aggregator and an end-point evidence monitoring system. We will develop and implement the two novel components and demonstrate their capability to verify components’ operational security and enhance network visibility. The outcomes are expected to provide important preliminary results for a planned large proposal submission.

10. Title: Coding theory for modern secure communication and computation
PI: Gretchen Matthews
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Research Program

Summary: This project uses coding theory to address two key themes in connected computing, modern communications and computations: providing noise control and mitigation in support of quantum computing and ensuring security and efficiency of quantum-safe schemes for encryption. Objectives include the design of new quantum error correcting codes which mitigate errors in qubits and updated security analysis of quantum-safe cryptosystems.

11. Title: Resilient cordinated control of renewable-based power grid under cyberattacks
PI: Ali Mehrizi-Sani
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Research Program

Summary: The overarching objective of this research proposal is to create and evaluate cyberattack-resilient algorithms for control of the renewables-based power grid. The United States is investing in generating electricity from renewable sources such as solar and wind. These resources differ from conventional thermal and hydro generation plants, which are based on large synchronous generators, in how they are connected to the grid: Instead of direct connection, they use power electronics devices called inverters to condition the generated electric power to have the voltage and current waveform required by the grid. This mode of generation introduces several challenges in power system operation and control. Examples of these challenges include smaller size, larger number, lack of mechanical inertia, and sensitivity to overcurrents and overvoltages. These present a major departure in both steady-state and dynamic characteristics of the power system. Addressing these challenges requires closer coordination between IBRs through communication- based distributed control. However, communication introduces the risk of cyberattacks. Therefore, this proposal focuses on creating algorithms that are resilient-by-design to cyberattacks. We argue that this line of research addresses a critical gap as we are at a point when different technology areas are maturing and have introduced new applications: these technologies are power system IBR-based edge devices, communication (5G and beyond), and availability of control and computational power. Upon its successful completion, this proposal will enable the foundational work leading to creation of a start-up marketing cybersecurity solutions for the power grid.

12. Title: Bolstering safety and trust in multimodal foundation models through risk, attack, and defense analysis
PI: Chris Thomas
Lead Institution: Virginia Tech
Co-PIs & Institution: Ruoxi Jia 
Funding Program: FY25 Research Program

Summary: Recent years have seen an explosion of interest in large language model AI systems like ChatGPT. Despite the promise of these models, their use comes with many risks, such as producing harmful outputs or being compromised by malicious actors. These risks have limited the deployment of such models in critical settings where reliability and security are important. Advanced AI systems are now moving beyond purely textual models to handle multiple types of data, such as both text and images, which introduces new risks. This project aims to make the next generation of large multimodal models, like GPT-4o, safer and more reliable by understanding the unique risks associated with these models and creating new methods to test and improve their security. It will develop techniques to detect and fix problems in the data used to train these models and spot hidden threats that could cause unexpected behaviors. The goal is to ensure these AI systems can be safely used in important areas like healthcare by enhancing their safety and reliability. Expected outcomes include improved techniques for testing AI safety, identifying risks, and ensuring models are safe to use. The project will share its findings through workshops and academic papers to help others understand and address AI security challenges.

13. Title: Privacy-Preserving Digital Twin Generation for Electronic Health Records with Large Language Models
PI: Xuan Wang
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Research Program

Summary: This project aims to develop high-fidelity digital twin generation methods for patient electronic health records (EHRs) to ensure personalized care in complex clinical settings while preserving privacy. Digital twins are anonymized and synthetic representations of real patients, providing valuable insights for healthcare research and analysis while minimizing clinical risks and reducing costs. However, generating digital twins for EHRs poses significant challenges due to data inconsistencies, diverse events, and model generalizability. To address these challenges, we propose novel methods combining temporal augmentation, external supervision, and multi-agent large language models to enhance digital twin accuracy and clinical validity. Additionally, we propose to demonstrate adaptability and effectiveness of our methods in real-world settings and test generalizability across various medical conditions and treatment procedures. This project will have a profound impact on healthcare and society, contributing to improved patient care, enhanced healthcare research, increased patient privacy, reduced healthcare costs, and broader applications in medical device development, pharmaceutical research, and public health policy. 

14. Title: Hidden Within – Phase II
PI: Dr. Agnieszka Miedlar
Lead Institution: Virginia Tech
Co-PIs & Institution: Janet Biggs, Paul Cazeaux, Tanner Upthegrove, VirginiaTech
Funding Program: FY25 Research Program

Summary:  Building on our large-scale video installation Hidden Within, produced with support from the CCI CyberArts and Design Program (2023-24), our interdisciplinary team—visual artist and Co-PI Janet Biggs, vision-impaired dancer Davian Robinson, mathematicians PI Agnieszka Międlar and Co-PI Paul Cazeaux, and sound specialist and Co-PI Tanner Upthegrove continues to push the boundaries of research and creative experimentation, drawing inspiration from steganography. At its core, our project explores how information is concealed, detected, and revealed. Beneath the surface, it delves into introspection, examining the many ways hidden information operates - whether as a tool for amplifying voices suppressed by oppression or as a means of spreading and uncovering misinformation.

15. Title: UVA Wise Global Engagement Program
PI: Karen Carter
Lead Institution: UVA - Wise
Co-PIs & Institution: None
Funding Program: FY25 Global Engagement Program

Summary:  The UVA Wise Global Engagement program initiative will support two objects: 1) establishing a research framework in artificial intelligence (AI) and machine learning (ML) at UVA Wise and 2) providing experiential learning opportunities to work collaboratively with one or more of UVA Wise’s international sister institutions. The impact of these objectives is especially important as the development and adoption of new and emerging AI and ML technologies have become omnipresent. The development of a global ecosystem lens will help to inform what steps are currently underway and which steps are needed as the world navigates AI’s and ML’s influence upon cybersecurity and data protections. This collaboration will provide an initial foundation for future exchanges of research, faculty, and students for UVA Wise and its sister international institutions. Generating not only an alliance within the AI and ML research development but an expectation of continued collaborative involvement as the global community continues to gain awareness and understanding of these technologies with an exploratory analysis focus to create a science-backed body of evolving, shared work. 

16. Title: Using Honeypots to Understand International Differences in Cybercrime
PI: Dr. Thomas Dearden
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Global Engagement Program

Summary:  This research proposal focuses on addressing cybercrime through an interdisciplinary approach that combines social sciences with cybersecurity. The project, led by Dr. Dearden, aims to understand the human factors behind cybercrime, particularly the geographic and socioeconomic influences on cyber offenders' behavior. The research team has already conducted studies using honeypot data to examine international differences in breach attempts, finding correlations between educational expenditures, inequality, unemployment, and cybercrime ac􀆟vi􀆟es. Their current work extends to analyzing offender behaviors post-breach and exploring methods to detect the relative location of individuals using proxies. The project aligns with the Commonwealth Cyber Initiative’s mission by focusing on human factors and global security in society. It aims to prevent cybercrime by addressing root causes rather than just reacting to attacks. The researchers plan to produce peer-reviewed articles, present at conferences, and submit additional grant proposals.

17. Title: Brazil GEOINT Initiative Proposal
PI: Dr. Andrew Foy
Lead Institution: Radford University
Co-PIs & Institution: Prem Uppuluri, Steve Childers, Radford University
Funding Program: FY25 Global Engagement Program

Summary:  Through this proposal Radford University will work with SENAI in Recife Brazil to offer undergraduates, graduates in the local universities (such as UFPE – Federale University of Pernambuco) as well as professionals in SENAI, a one weeklong workshop focused on geospatial intelligence (GEOINT) activities, paired with a proposed cybersecurity and entrepreneurship workshop during Summer 2025. The workshop will use the RU Geospatial Portal and GEOINT ecosystem to effectively provide just in time learning, exploitation and analysis of imagery and geospatial information to describe, assess, and visually depict physical features and geographically referenced activities on the Earth. Workshop participants will work on business ideas related to GEOINT and will implement best cyber security practices in collaboration with the researchers at SENAI.

18. Title: Virginia International Research on Technology and Unified Cybersecurity in Electric Systems (VIRTUE)
PI: Dr. Ali Mehrizi Sani
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Global Engagement Program

Summary:  Summary. The objective of this seed funding request is to signi_cantly advance Virginia Tech's power grid cybersecurity research and global leadership position through several planned activities with the University of Hong Kong (HKU). The VT team works on the cybersecurity of the sustainable power grid, e.g., the modernized power system that includes a host of inverter-based resources (IBR) for renewable generation with 5G coordination. The HKU team works on the niche area of resilient IBR-intensive energy systems with various uncertainties. This leads to a unique synergy for energy applications: cybersecurity of the uncertain power grid, where uncertainty can be due to multiple factors: (i) generation; (ii) planning and expansion, and (iii) communication channel. The PI has a long history of successful international collaboration, with many that turned into new research areas and supported through external funding agencies, including NSF and European Union. The planned activities include (i) collaborative research visit (including a talk by the VT PI and planning and preliminary studies to provide a clear PhD-level project plan for students who will be working on the project); (ii) design and delivery of a workshop to evolve into a two-day industry short course o_ering through VT Continuing Education; (iii) joint publications; and (iv) submission of a major proposal. The expected impact of this project on the Commonwealth is two-fold: (i) build world-leading research and innovation capabilities; (ii) help close the cybersecurity workforce gap by preparing students for the job market.

19. Title: The 1st Workshop on Agentic AI for Science: Hypothesis Generation, Comprehension, Quantification, and Validation
PI: Dr. Dawei Zhou
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Global Engagement Program

Summary:  Our workshop mission is to foster interdisciplinary collaboration to develop fully autonomous AI systems, addressing challenges like benchmark datasets, human-AI collaboration, robust tools and methods for validating AI outputs, and trustworthiness. By tackling these issues, we can unlock AI’s transformative potential in research. In this workshop, themed Agentic AI for Science, we will explore these critical topics and welcome diverse perspectives. We will focus on integrating agentic AI systems to enhance scientific discovery while upholding rigorous standards. For AI to contribute effectively, it must generate novel hypotheses, comprehend their applications, quantify testing resources, and validate feasibility through well-designed experiments. This workshop serves as a vital forum for collaboration and knowledge-sharing aimed at redefining the landscape of scientific discovery.

20. Title: Cyber-Attack Resilient Distributed and Explainable AI with Zero Trust Architecture
PI: Dr. Zeb Bowden
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Cybersecurity for Al and AI for Cybersecurity Program

Summary: This research project will create novel theories and techniques for constructing stakeholder-centric, secure data-sharing and analytics systems; specifically, our team aims to devise algorithms and theories that integrate federated learning and AI to enhance the security of data analysis in big data collaborative supply chains composed of several stakeholders constrained by data-privacy and confidentiality requirements. In this project, we intend to leverage distributed AI through Federated Learning (FL) by developing a novel privacy-preserving technique that enables rigorous, secure collaboration. In this sense, we intend to use a zero-trust architecture to strictly validate global model updates while ensuring continuous protection for local models. An advanced Explainable AI (XAI) system will autonomously monitor, predict, and detect potential attacks, facilitating root cause analysis and dynamic adjustment of learning parameters.

21. Title: Towards a knowledge-guided Foundation Model for Longtail Malware Detection
PI: Dr. Bo Ji
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Cybersecurity for Al and AI for Cybersecurity Program

Summary: In this project the PI will take the responsibility of using ns3 simulator to generate simulated network traffic data and discuss research ideas at weekly meetings with PIs . Perform network simulations to generate realistic network traffic data that will be used to overcome data imbalance problem, assist with the creation of a comprehensive network traffic benchmark dataset for pretraining the proposed foundation model and work with PIs at William & Mary to prepare annual report and final report.

22. Title: Intelligent and Secure Wireless Sensor System for Monitoring Cattle on Farms
PI: Dr. Sook Shin
Lead Institution: Virginia Tech
Co-PIs & Institution: None
Funding Program: FY25 Cybersecurity for Al and AI for Cybersecurity Program

Summary: The rapid integration of AI-based monitoring systems in smart farms offers tremendous potential for improving livestock management, particularly in disease detection and animal welfare. However, these systems face significant biosecurity challenges, such as AI vulnerabilities, cross-farm contamination, and targeted bioterrorism. Inspired by Nassim Taleb’s concept of anti-fragility, this research aims to develop a robust, antifragile AI-driven wireless sensor system that can adapt, evolve, and grow stronger under high uncertainty, dynamics, threats, and failures to ensure early detection of diseases and prevent or mitigate pathogen spread. To achieve this, we propose a framework called FragileGuard, an antifragile AI-driven wireless sensor system designed for accurate monitoring and early disease detection. In this context, we will focus on fortifying smart farm environments, especially for cattle, by creating secure AI models that withstand cyber and adversarial threats, safeguarding both animal health and the broader food supply chain. Building on these goals, integrating AI-driven monitoring systems is about leveraging technology for disease detection and addressing the pressing biosecurity risks such as AI vulnerabilities, contamination across farms, and bioterrorism threats. By embracing the anti-fragility principle, this research is geared toward developing adaptive AI systems that thrive under dynamic and uncertain conditions, ultimately reinforcing the resilience of smart farms.

23. Title: Software Defined Radio and O-RAN for Mobile Distributed MIMO (dMIMO)
PI: Lingjia Liu
Lead Institution: Virginia Tech
Co-PIs & Institution: none
Funding Program: FY25 Security Scholars

Summary: Massive MIMO is an enabling technology for 3GPP 5G NR and is envisioned as a key one for Beyond 5G networks (e.g., 5G-Advanced, 6G, etc). To circumvent the form factor limitation of gNBs and to ensure a more unified network performance, distributed MIMO (dMIMO) operations such as distributed antenna systems (DAS), dynamic point selection (DPS), coordinated multi-point (CoMP) (a.k.a. network MIMO), and multiple transmit/receive point (multi-TRP) have been introduced in various releases of 3GPP. In the “Mobile Distributed MIMO: Learning Meets Spreading in Networking” project funded by OUSD(R&E)’s FutureG Program, we will be mainly focusing on system design and hardware prototype for mobile dMIMO systems. In this CCI project, we will be focusing on the software defined radio (SDR) prototype of dMIMO as well as the O-RAN system design and prototype for mobile dMIMO systems.

24. Title: Graduate Research & Innovation Design: Secure Connectivity for Renewable Devices in the grid (GRID-SECURED)
PI: Ali Mehrizi-Sani
Lead Institution: Virginia Tech
Co-PIs & Institution: none
Funding Program: FY25 Security Scholars

Summary: The electric energy delivery system, commonly known as the power grid, is a critical infrastructure and a component of the national security.  Transition to renewable energy resources, which can be deployed in a distributed manner close to power consumption sites, has necessitated the availability of communication infrastructure. Such infrastructure additionally has the potential to enhance the performance of control and protection facilities in the power grid. However, this also makes the power grid an attractive target for cyber-attacks. The overarching objective of this project is to create and demonstrate the GRID-SECURED framework. GRID-SECURED is a scalable, distributed, and cost-effective technology to enable secure and resilient communication to control and monitor distributed energy resources (DER), including grid-forming (GFM) inverters, within the power system and/or microgrids.

25. Title: Large Language Model for Access Control Policy Generation in Microservices or Cloud Computing 
PI: Peng Gao
Lead Institution: Virginia Tech
Co-PIs & Institution: none
Funding Program: FY25 Security Scholars

Summary: The project analyzed the performance of different LLMs across the key stages of access control policy generation, including pre-processing, policy identification, rule extraction, transformation, and verification.  Additionally, the team observed the prevalence and types of hallucinations in generated access control policies (e.g., incorrect rule generation, assumed knowledge of the correct frame type, inconsistencies in identifying temporal and location arguments, etc.) to establish a benchmark and propose a framework for addressing these limitations.