- Home |
- Overview of Projects for Summer 2025
Overview of Projects for Summer 2025
Privacy-preservation in Spectrum Sharing – Dr. Vini Chaudhary
Wireless communication relies on radio frequency (RF) spectrum to transmit and receive signals. This RF spectrum is a limited and finite resource, creating challenges in supporting the ever-growing number of users in next-generation wireless networks. Spectrum sharing is one of the potential solutions to this problem, which maximizes spectrum utilization by either dynamically or statically allocating the same frequency bands to multiple users/services. However, this approach faces significant cybersecurity challenges, including the need to detect interference between the users, presence of anomalous users, and privacy concerns while sharing spectrum data. This project will familiarize undergraduate students to these challenges, help them in performing research to overcome these cybersecurity issues faced during shared spectrum management, and train them to conduct field experiments to validate their research/solutions.
The primary objective of this project is (i) to develop and evaluate machine learning (ML) algorithm that can detect interfering and anomalous users’ signals that fully/partially-overlap in time and frequency domains (ii) to use homomorphic encryption-based training of the ML algorithm for preserving privacy of users’ data. The methodology includes:
- Hypothesis Development: The students will first create a synthetic dataset in which 2-3 type of RF signals (e.g., 4G/5G, spread spectrum as anomaly, and radar) will be generated using standard-compliant MATLAB toolboxes and arranged in time-frequency domains so as to represent interference scenarios (full/partial overlaps, hidden anomaly signals). Thereafter, they will use state-of-the-art neural networks (e.g., Alexnet NN, etc.) to detect interference and anomalous RF signals. Next, the homomorphic encryption-based training will be explored and applied to this NN. It leverages homomorphic encryption, which allows computations to be performed directly on encrypted data without needing to decrypt it first, thereby ensuring reverse engineering back to the users’ RF signals becomes increasingly difficult.
- Experimental Design and Real-world Dataset Collection: Students will learn how to use software-defined radios (SDRs) to transmit and receive the above-mentioned RF signals (e.g., National Instrument’s USRP X310) using my previously designed open-source application programming interface (API) for RF dataset collection in collaboration with National Instruments and Northeastern University. They will generate a variety of fully/partially overlapping and hidden signal scenarios. These experiments will be structured to rigorously test the hypotheses. This real-world data will inherently capture wireless channel and transceiver impairments in the received signals.
- Analysis: The real-world collected data will be used to analyze the effectiveness of proposed solution in detecting interfering and anomaly users’ signals in shared spectrum context while preserving the privacy of spectrum data. Since this dataset will capture real-world artifacts, it will enable assessing the viability of the proposed cybersecurity solution in a real shared spectrum context.
Students will develop the following research skills through this project: (i) reading and understanding research papers, (ii) designing experiments, (iii) collecting datasets, (iv) writing new codes and adapting existing codes as per requirement, (v) analyzing results, and (vi) writing technical papers. To help them in learning these skills, there will be weekly meetings to brainstorm ideas, discuss issues faced by students, the strategies to overcome these issues, and student presentations (at a later stage). This will ensure that the students will develop critical thinking and will be able to progress and complete the project in the stipulated amount of time. The project will help student in acquiring technical programming skills in Python/MATLAB for implementing privacy-preserving NN for detecting interference and anomalous RF transmissions in shared spectrum applications. The students will get a hands-on experience in using SDRs and the RF dataset collection APIs to transmit and receive RF signals. This will enable them to collect real-world datasets not just for this project, but also in their future projects. Students will also get familiarized with different representations of RF signals, e.g. spectrograms, power spectral densities, etc., which will enable them to visually understand how signals and their interference will look in the time-frequency domains. These skills will enable students to carry out research effectively.
In this project, the students will majorly contribute to developing interference and anomaly detection algorithm in shared spectrum use-cases while preserving privacy of users’ data used in these algorithms. The students will generate synthetic dataset as an initial start point, design algorithm using Alexnet NN, write code-base in Python for including homomorphic encryption-based training, analyzing results on synthetic dataset. After this, students will design a real-world experiment using SDRs and collect dataset for the interfering and anomaly signals in shared spectrum cases. Next, they will analyze results of the designed algorithm on real dataset, draw conclusions, present those results in meeting, and write research paper to wrap up the project. Throughout the project, students will comprehensively document each and every steps. The main advantage of this project is that it will develop experimental, technical, and analytical skills of the student. The hands-on experience in using SDRs to set up experiments involving RF communication signals can be used in any other project related to using ML in next-generation wireless communication networks. The student will primarily work independently on focused aspects of the project, such as algorithm development and data analysis, while also collaborating with other team members to ensure seamless integration into the overall project. Regular group meetings, facilitated through Microsoft Teams/Webex, will offer valuable opportunities for the student to enhance collaboration skills by engaging with peers in related research areas. This balanced approach enables the student to develop deep expertise in a specific area while broadening their technical knowledge through networking and teamwork.
Cybersecurity for AI Robotic Systems – Dr. Jingdao Chen
Robotics, automation, and related Artificial Intelligence (AI) systems have become pervasive bringing in concerns related to security, safety, accuracy, and trust. These include robotic systems such as autonomous cars, medical robots, and recreational drones that operate in our households and workplaces. With growing dependency on physical robots that work in close proximity to humans, the security of these systems is becoming increasingly important to prevent cyber-attacks that could lead to privacy invasion, critical operations sabotage, and bodily harm. The current shortfall of professionals who can defend such systems demands development and integration of cybersecurity tools. This project will study current trends in robotic cybersecurity and train undergraduate students through the REU program to understand threats and vulnerabilities of AI robotic systems and perform research to defend against cyber-attacks on these systems
One specific topic on which an REU student would work on is adversarial attacks on vision-language navigation systems for robots. Research objectives include (i) perform a comparative evaluation of various prompt injection attacks on vision-language navigation systems (ii) evaluate the drop in performance of robot navigation systems as a result on adversarial attacks (iii) determine the effectiveness of defense strategies such as ensemble models on adversarial attacks. Students will use
state-of-the-art vision-language navigation systems available as open-source code such as NavGPT, LM-Nav, and VL-Map as a starting point for code development. Students will then implement various adversarial attacks such as prompt injection and adversarial inputs in addition to defense strategies such as ensemble models. Students will develop a hypothesis on the research objectives, design experiments to evaluate the hypotheses, write code and gather data to analyze the outcome of adversarial attacks as well as the effectiveness of defense strategies.
Students’ involvement in research activities will be focused on development of research skills such as experimental design, data analysis, and scientific writing. Experimental design and critical thinking skills will be taught and reinforced through weekly research discussions and brainstorming sessions. Scientific writing and communication skills will be honed through paper reading assignments, weekly presentations, and technical report writing. The research tasks assigned will allow the students to grow in technical abilities important in computer science research such as programming, package management, and data analysis. Students will be guided on using specific software tools such as ROS for robotic code development, PyTorch for deep learning, and Github for version control.
The REU students will mainly contribute to security algorithm design, data analysis, and development of a defense system for alleviating adversarial attacks on AI robotic systems. Roles and responsibilities include developing navigation models using Python and PyTorch on multiple platforms such as the Matterport 3D simulator and a physical Ghost Robotics Vision 60 unit. In addition, the students will use a test set to calculate overall algorithm performance metrics, including navigation error, success rate, and trajectory length. Even though students will mainly work independently on focused research thrusts, students will have the opportunity to develop collaboration skills through meetings with other students working on similar areas. Group meetings will be held on a regular basis and group work will be coordinated through Microsoft Teams. This allows students to acquire depth in a particular research area while maintaining breadth of technical knowledge by networking with peers.
Explainable Malware Intrusion Detection based on Artificial Immune Systems – Dr. Sudip Mittal
Intrusion detection is a complex process that involves dealing with malicious individuals who use a variety of techniques, from technical skills to social engineering. This process often involves deceptions and misdirection, where things may not be as they seem. Due to this complexity, there is a need for algorithms that can handle imprecision, uncertainty, and approximations. Another important need in today's cybersecurity landscape is the explainability of AI models, especially in the context of Malware Intrusion Detection. Explainable AI (XAI) systems in cyber defense can significantly enhance organizational cybersecurity operations. This project will leverage Soft Computing and Computational Intelligence, particularly Artificial Immune Systems (AIS), which are inspired by the Human Immune System (HIS), to develop explainable malware intrusion detection systems. Just as the Human Immune System identifies and protects the body against foreign pathogens, AIS can be used to detect and protect against malware. AIS has been applied in network protection in the past, and with recent advancements in AI, it holds promise for malware analysis and intrusion detection. A significant advantage of using AIS-based approaches is their inherent explainability, making them suitable for the development of XAI-based cybersecurity solutions.
This project aims to create an explainable malware detection and diagnostic system based on Artificial Immune System Danger Theory (AISDT). The system will integrate subsystems to monitor and collect relevant information, along with an explainable computational model to aid in diagnostic analysis. The design will prioritize modularity, scalability, adaptability to changing scenarios, and resource efficiency.
The project is divided into three main tasks:
- Survey and Data Collection: Students will start by reviewing the literature on soft computing approaches and malware detection, focusing on algorithms like genetic algorithms, neural networks, evolutionary algorithms, and swarm intelligence. They will study the concepts of HIS, AIS, and AISDT. Simultaneously, students will engage in data collection, even though the PI already has access to the necessary data from previous research. This process will help students gain a deeper understanding of data collection and its importance in modeling malware detection.
- Data Modeling and Evaluation: Using the data collected in Task 1, students will develop malware detection models based on various AIS approaches. They will compare the performance of these models and explore the theoretical implications of extending AISDT to malware intrusion detection. During this phase, students will also learn about traditional intrusion detection metrics and AI evaluation strategies, deepening their understanding of both the algorithms and their practical applications.
- Explainability Analysis: Students will focus on understanding the importance of explainability in cybersecurity. They will learn how the explainability of AI-driven cyber defense systems can improve operations in a Security Operations Center (SOC). The project will introduce students to key concepts of AI explainability and modern techniques such as Layerwise Relevance Propagation (LRP) and local interpretability. A significant portion of this task will involve evaluating the explainability of the models developed in Task 2
Throughout the project, students will develop a range of research skills essential for cybersecurity, including literature review, data collection, and the development and evaluation of AI models. They will gain experience in understanding and applying sophisticated soft computing techniques to real-world cybersecurity problems. Additionally, students will enhance their technical programming skills, particularly in Python, as they implement and test their malware detection models. The project will also develop students' abilities to critically evaluate the performance and explainability of AI models, preparing them for future research in AI and cybersecurity. Regular meetings will be held to discuss progress, address challenges, and brainstorm ideas, ensuring that students stay on track and develop the necessary critical thinking skills to complete the project successfully.
Students will play a key role in each phase of the project, starting with the literature review and data collection. They will then move on to developing and testing malware detection models using AIS-based approaches. As they progress, students will analyze the performance of these models and focus on understanding and improving their explainability. Throughout the project, students will document their work carefully, present their findings in meetings, and contribute to writing research papers summarizing the project's outcomes. The hands-on experience with cutting-edge AI techniques and the focus on explainability will equip students with valuable skills for future careers in cybersecurity and AI research. Collaboration with peers in related research areas through regular group meetings will also enhance their teamwork and communication skills.
Moving Target Defense Against Cyber-Attacks – Dr. Charan Gudla
This project explores Moving Target Defense (MTD) as a cybersecurity strategy to mitigate cyber threats by dynamically mutating IP addresses of network hosts. Attackers rely on reconnaissance and persistent access to exploit vulnerabilities; MTD disrupts this process by continuously altering network addresses, making tracking and exploitation significantly harder. The research focuses on developing and evaluating various IP mutation strategies to enhance security while minimizing disruptions to legitimate network activities. Using Mininet for simulation and Software-Defined Networking (SDN) controllers like Ryu, different mutation algorithms will be tested against various attack vectors to assess their effectiveness. Key performance metrics, such as attack prevention success rate, network stability, and latency impacts, will be analyzed. Collected data will be processed using statistical methods and machine learning to determine the optimal mutation strategies for modern network environments. By creating an adaptive, automated MTD framework, the project aims to improve cybersecurity resilience against reconnaissance attacks, unauthorized access, and persistent threats. The findings will contribute to the advancement of dynamic defense mechanisms in emerging network infrastructures, providing a robust security layer for next-generation computing and communication systems.
The primary objective of this project is to develop and evaluate algorithms that dynamically mutate the IP addresses of network hosts as a defense mechanism against cyber-attacks. The student will focus on creating a prototype system that implements various IP mutation strategies and assesses their effectiveness in preventing unauthorized access while minimizing disruptions to legitimate network activities. The research methodology includes:
- Hypothesis Development: The project will begin with the formulation of hypotheses regarding the effectiveness of different dynamic IP mutation strategies in mitigating cyber-attacks under various network conditions.
- Experimental Design: A series of experiments will be designed and implemented using network simulation tools such as Mininet, and within controlled Software-Defined Networking (SDN) environments like those managed by Ryu. These experiments will be structured to rigorously test the hypotheses and explore how different IP mutation algorithms perform against a range of attack vectors.
- Data Collection: Throughout the experiments, comprehensive data will be collected on key performance metrics, including the effectiveness of attack prevention, the impact on network stability, and any latency introduced by the IP mutation processes. This data will be critical in assessing the viability and efficiency of each mutation strategy.
- Analysis: The collected data will be analyzed using advanced statistical methods and, where applicable, machine learning techniques. This analysis will aim to identify the optimal IP mutation strategy that balances security enhancement with minimal disruption to legitimate network traffic, ultimately contributing to the development of a robust and adaptive network defense mechanism.
Throughout the project, the students will develop research skills crucial for cybersecurity. They will begin by designing algorithms capable of dynamically mutating IP addresses to thwart cyber-attacks and learning how to adapt and optimize these algorithms in real-time network environments. The students will also gain hands-on experience in experimental design, where they will test the efficacy of different IP mutation strategies under various attack scenarios and network conditions. As part of their work, they will collect and analyze network performance data, including metrics like latency, security effectiveness, and system stability, to determine the most effective approaches. Additionally, the students will refine their scientific writing skills, documenting their methods, analyses, and conclusions in a clear and structured manner. Through this process, they will enhance their critical thinking abilities, particularly in evaluating the trade-offs between security enhancements and potential disruptions to network operations, ultimately identifying strategies that best balance these considerations. The students will also acquire critical technical skills essential for implementing and evaluating advanced cybersecurity strategies. They will gain proficiency in programming by writing Python scripts to automate network experiments and manage the dynamic IP mutation algorithms central to the research. The students will also develop expertise in network simulation using tools like Mininet, where they will simulate complex network environments and rigorously test the effectiveness of various IP mutation strategies. Additionally, the students will work with Software-Defined Networking (SDN) controllers, such as Ryu, to implement and manage dynamic IP address changes across the network, ensuring that the mutation processes are both secure and seamless. These technical skills will be integrated into the research activities, providing the student with hands-on experience in cutting-edge network security technologies.
In this project, the student will take on a vital role in developing and evaluating a dynamic IP mutation algorithm designed to enhance network security against cyber-attacks. Their responsibilities will include designing and implementing key components of the algorithm, focusing on how to effectively and securely mutate IP addresses in real time. The student will conduct experiments to assess the performance and effectiveness of these algorithms under various attack scenarios and network conditions, meticulously analyzing the resulting data to identify the most effective mutation strategies. Additionally, they will be tasked with thoroughly documenting the entire research process, including the methodologies used, the results obtained, and the conclusions drawn, culminating in a comprehensive research report. Through this hands-on experience, the student will contribute significantly to the project's goals while developing their own technical and analytical skills. The student will work independently on specific aspects of the project, such as algorithm development and data analysis while collaborating with other team members to integrate their work into the overall project. Regular meetings will be held to ensure the student is on track and their contributions align with the overall research goals.
Mitigating Jamming Attacks in Wireless Networks – Dr. Maxwell Young
Wireless networks are vulnerable to malicious devices that deliberately disrupt the shared communication medium; this is known as jamming. Over the past decade, jamming attacks have evolved from a mostly theoretical risk into a credible threat against wireless systems. Mitigating these attacks requires knowledge of wireless technology and standards, security threats, and algorithm analysis. This project directly addresses an important aspect of cybersecurity in emerging technologies by addressing the design of defenses with provable security guarantees against jamming.
This project focuses on evaluating the effectiveness of recent results on medium access control (MAC) algorithms that are designed to withstand jamming attacks in wireless networks. Utilizing MATLAB, the project involves setting up a simulated wireless network environment where different types of jamming attacks can be created and their impact on MAC algorithms can be evaluated. The REU students will learn several jamming-resistant MAC algorithms. Using MATLAB, the students will simulate a wireless network environment where they can implement and test various types of jamming strategies (i.e., constant, random, and reactive). Within this simulation, the students will implement the contention resolution algorithms. Data will be collected on key performance metrics, such as throughput, latency, packet delivery ratio, and collision rates. They will use statistical tools to analyze this data, looking to identify which algorithms are most effective at mitigating the impact of jamming. The culmination of their work will be a comprehensive report that details their methodologies, findings, and conclusions.
Students will develop a range of research skills. They will hone their abilities in experimental design by setting up and configuring simulations, allowing them to understand the impact of attacks on wireless communication. Data analysis skills will be enhanced as students learn to gather, process, and interpret large sets of data to evaluate the effectiveness of different algorithms, using statistical methods to discern patterns and outcomes. Students will also gain experience with scientific writing through the preparation of a project report that conveys the technical content in a clear and structured manner. Critical thinking will be emphasized as they must assess the results, theorize on the implications, and suggest improvements or new areas of study based on their findings. The students will also develop technical programming skills for scripting in simulation environments using MATLAB. This includes setting up and conducting network simulations, where they'll configure parameters, implement jamming scenarios, and analyze network behaviors. Additionally, students will gain expertise in statistical analysis by employing MATLAB and R to analyze data from their simulations. Skills in statistical testing and data visualization will be important for evaluating the effectiveness of various algorithms and describing the main findings. Finally, these findings will be presented in a document LaTeX for academic and technical reporting.
A primary objective of this project is to ensure that each student gains a comprehensive understanding across all relevant areas, including algorithm design, simulation of wireless communications, understanding of jamming strategies, data analysis, and technical writing. Students will be grouped to specialize in different aspects of jamming-resistant MAC algorithms, allowing them to delve deeply into specific algorithmic details. Additionally, as simulation designers, they will model wireless communications using an abstract time-slotted channel model that incorporates simple ternary feedback (indicating no transmission, a single transmission, or multiple transmissions) and established path loss effects. Different groups will explore and implement various path loss models to understand their impact on communication. Furthermore, each group will analyze the effectiveness of these algorithms using statistical methods, but individual students will take responsibility for analyzing and documenting the results for distinct jamming strategies and performance metrics. This structured approach ensures that while each student focuses on and is responsible for specific tasks, they also gain skills across a broad spectrum of the project’s scope.
As described above, students will work collaboratively in groups to explore different aspects of the project, such as algorithm design and simulation setup, and data analysis. Within these groups, individual students will take responsibility for specific tasks, like implementing different path loss models, analyzing distinct jamming strategies, and calculating performance metrics. Regular group meetings will facilitate the integration of individual work. This approach ensures that while students develop specialized skills independently, they also contribute to the overall project goals.