48440 research outputs found
Sort by
Multiparty Computation Schemes: Physical Security and Applications in Secure Implementations
No full textThe increasing reliance on cloud-based computation, machine learning services, and collaborative design tools has introduced significant challenges in ensuring data confidentiality and computational integrity. Although Secure Multiparty Computation (MPC) and Fully Homomorphic Encryption (FHE) offer strong theoretical guarantees, their practical adoption remains limited due to performance bottlenecks and exposure to physical attacks such as side-channel leakage and fault injection. This dissertation addresses these challenges through a set of hardware-assisted secure computation frameworks that improve both efficiency and robustness. GarbledEDA introduces a privacy-preserving electronic design automation solution that secures intellectual property during hardware verification using optimized garbled circuits. GuardianMPC builds on this foundation by accelerating secure neural network inference through parallel garbled circuit evaluation and customized hardware modules for oblivious transfer, while also incorporating backdoor detection mechanisms to ensure model integrity. To highlight practical vulnerabilities, we present Goblin, a timing side-channel attack that targets widely-used MPC frameworks, revealing how variations in execution time can be exploited to recover secret inputs. FaultyGarble demonstrates that laser fault injection can extract proprietary neural network parameters from garbled circuit-based secure inference systems, exposing the limitations of cryptographic guarantees in the presence of physical adversaries. We also present Bake It Till You Make It!, a novel temperature-based side-channel attack that shows how controlled heating can bypass masking defenses in secure hardware implementations. To address such risks, we propose HWGN2, a secure inference framework based on secure function evaluation, which is designed to resist power, timing, and electromagnetic side-channel attacks. Finally, Garblet introduces a chiplet-aware architecture for secure MPC deployment across heterogeneous hardware platforms. By distributing garbled computations and integrating hardware-level optimizations, Garblet reduces communication overhead and maintains strong security, even when operating across untrusted components. These contributions demonstrate that secure computation in modern systems requires a multi-layered approach, combining cryptographic techniques with physical defenses and architectural awareness. This dissertation advances the state of the art by bridging the gap between theoretical protocols and practical, secure implementations that can withstand real-world adversaries and physical threats
Lithium and Transition Metal Recovery from Spent Lithium-ion Batteries
No full textThe growing demand for lithium-ion batteries (LIBs) and rising lithium salt costs are placing increasing strain on the LIB supply chain. Recycling spent LIBs has emerged as a promising strategy to alleviate these pressures while mitigating environmental impact. Although much research has focused on recovering valuable transition metals, lithium recovery techniques remain underdeveloped. Traditional methods dissolve all metals in acidic media and sequentially precipitate transition metals before extracting lithium, often resulting in low efficiency and purity. Moreover, existing recycling strategies typically regenerate cathode materials with low nickel content, failing to align with the market’s shift toward Ni-rich compositions. To address these gaps, this research prioritizes lithium-first extraction using various organic acids to enhance lithium recovery efficiency and purity. Additionally, a scalable and economically viable hydrometallurgical upcycling process has been developed to regenerate Ni-rich cathode materials from mixed Ni-lean spent LIBs. This approach supports a closed-loop system that bridges older cathode types with next-generation materials, offering a sustainable and commercially competitive pathway for LIB recycling aligned with current industry needs
Neuronal Circuits Underlying Sensory Processing In C. elegans
No full textSensory processing is essential for all animals. When an external stimulus is sensed, the nervous system creates a corresponding behavior. This process typically starts with perception via a receptor. The signal created by activating the receptor is neuronally passed and processed by circuits of neurons, resulting in the appropriate behavior. Neural circuits that govern sensory processing can be exceedingly complex. Therefore, when studying neural circuits on an individual-neuron level, it is necessary to use a simpler model organism. In this thesis, investigation to untangle neural circuits governing sensory processing is performed in the model organism Caenorhabditis elegans. C. elegans are small, transparent nematodes that are amenable to behavioral and genetic research techniques. And most importantly, C. elegans contain a simple nervous system that shares homology with the human nervous system. Chapter one outlines sensory processing and how C. elegans can serve as a model for studying sensory processing. In chapter two, I explain how pheromones control behavior in worms. Throughout the thesis, these pheromones are used as a tool to study sensory processing. Chapter three is concerned with sensory receptors, specifically, receptors to a worm mating pheromone. Chapter four outlines principles underlying the sensory processing of competing sensory cues. And finally, chapter five examines synaptic proteins that are involved in modulating avoidance behavioral responses. Together, this body of work outlines various aspects of neuronal sensory processing
Neurological and Psychosocial Effects of Habitual Nicotine Use
No full textNicotine addiction has been on the rise since e-cigarettes were introduced to the US market in 2007. However, there is not much known about the consequences of chronic nicotine exposure in young adults. This interdisciplinary project utilizes a biopsychosocial approach to understand these effects through a) experimentation with C. elegans and reproductive behaviors, b) development of a cue reactivity protocol using emerging EEG technology, and c) a psychosocial survey focused on vaping behaviors and mental health outcomes administered to 87 WPI students. This project found that C. elegans laid slightly more eggs after being exposed to nicotine after 24 hours and vapers showed more pronounced results for anxiety and depression
The Role of PLCβ in C. elegans Health and Neuron Function
No full textPhospholipase C beta 1 (PLCβ1) couples with G-proteins in the G-protein coupled receptor signaling pathway to primarily control calcium signaling. GPCR/Gαq signaling allows cells and organisms to communicate extracellular signals to intracellular compartments, serving as a main mode of communication with external environments. Interestingly GPCR signaling pathways are increasingly relevant for potential new drug avenues with over half of the FDA-approved medications targeting GPCR receptors. Since PLCβ1 is an important downstream effector of this pathway, understanding PLCβ1 and its binding partners could aid drug development, and novel insights may emerge for treating neurodegenerative disorders, such as Alzheimer’s disease, ALS, schizophrenia, Parkinson's disease, and autism spectrum disorder. This dissertation studies PLCβ1 and its roles in stress response, memory formation, and neuronal differentiation, using the simple organism C. elegans to explore these functions. When GPCRs are activated, cytosolic PLCβ1 moves to the cell membrane, leading to increased calcium influx while releasing binding partners involved in stress granule formation and RISC processing. This release influences stress granule formation, adversely affecting C. elegans health within their structured neural network. Additionally, RISC components, including TRBP and Ago2, and the transcription factor EGR-1, collaborate to enhance neuronal signaling related to differentiation and synaptic plasticity. The expression and location of these proteins are crucial for neuronal health, visible through changes in memory consolidation and stress granule dynamics relative to various PLCβ1 levels. Insights provided in this dissertation highlight the key effects of dysregulated PLCβ1 signaling in C. elegans providing novel mechanistic understanding of downstream GPCR signaling
Sensing Through Faults: Collective Perception by Imperfect Robot Swarms
No full textSwarm robotic systems present many advantages in helping us solve important challenges of today, from planetary exploration to the monitoring of hazardous zones. A fundamental problem that exists in such applications is the problem of collective perception, in which the swarm must reach a consensus on a coherent representation of the environment. An important niche in this problem is that of swarms formed by minimalistic individuals, defined by the severely limited capabilities in terms of storage, computational capabilities, and communication bandwidth. Minimalistic robots often display high levels of sensory noise, which further complicates the construction of coherent shared environment representations. However, past studies on the collective perception problem involve robots with perfect sensing or small numbers of faulty robots. Instead, this dissertation considers the problem of using swarm robots whose sensors are imperfect. I present a probabilistic algorithm that helps the robots collectively decide the frequency of an environmental feature. The algorithm, derived from optimal estimation techniques and a decentralized Kalman filter, enables the swarm robots to make accurate frequency estimates even with severe sensor degradation. Further, this dissertation studies the situation when the sensor degrades without the robots' knowledge. Specifically, I take two approaches to this issue: when the sensor degradation is static and when it is dynamic. In the former case, I introduce an adaptive self-calibration algorithm that leverages a hypothesis test to inform the correct moment for the robots to self-calibrate their sensor accuracy. In the latter case, I implement an extended Kalman filter to simultaneously track a robot's sensor accuracy and compute the frequency estimate (of the environmental feature). In both cases, my proposed approaches improve the estimation performance of a swarm than those without a mitigation strategy
Advancing Humanoid Robots: Development of Balancing and Assisted Walking Along With Improved Hardware
No full textThis project marks the latest iteration of WPI’s endeavor to design a reliable, open-source, 3D-printed humanoid robot, named Ava. Previous teams have worked on developing the motor initialization methods and assisted walking framework that our team utilized and further developed this year. Our team aimed to improve the assisted walking and self-balancing capabilities through sensor integration, control methods, and mechanical redesigns. We introduced new hardware and software systems, including a multiple Inertial Measurement Unit feedback system and newly redesigned waist parts. We transitioned from joint-space to task-space trajectory generation using Inverse Kinematics for Python, allowing for refined end-effector control to perform assisted walking. Our sensor control implementation allowed the robot’s upper body to dynamically balance while waving in simulation. These developments represent a significant step toward unassisted humanoid walking, laying the groundwork for more advanced bipedal behaviors in future iterations
Developing an Interactive Satellite-Based Environmental Education Program for Green Week: Year One
No full textIn response to the EU’s climate goals, the Romanian government founded a nationwide initiative called Green Week. The current framework lacks structure, putting excess pressure on teachers. Our team partnered with Pur și Simplu Verde to create a video game and other educational materials to support these teachers. We assessed current practices and identified gaps in environmental education. This led us to create a game prototype, an easy-to-navigate website, a database of essential information, and Green Week activity examples. Topics were based on NGO interviews and student input to maximize both interest and importance. By centralizing these resources and presenting topics in a game format, our project made Green Week programming more accessible and engaging
Advancing Resiliency and Preparedness in Rural Puerto Rico
No full textNatural disasters and related emergencies have lasting effects on mental, physical, and spiritual health that can persist for months or even years. Trauma-informed emergency plans and holistic health practices can minimize these impacts. This project aims to strengthen the Apoyo Mutuo Agricola y Rural (AMA) emergency preparedness plan and support the development of infrastructure for holistic health and medicinal practices. By collaborating with AMA volunteers and conducting interviews with community members and local farmers, we were able to identify gaps in the existing plan. As a result, we contributed to a more comprehensive emergency plan, established gardening infrastructure, and created data systems to enhance AMA’s communication and information management capabilities
UAV-UGV Collaboration
No full textUnmanned Aerial Vehicles (UAVs), specifically quadcopters, are increasingly used for autonomous tasks, but their limited battery life restricts operational duration. In low-infrastructure environments such as during disaster response or search and rescue operations, short battery life jeopardizes missions. To address this problem, we present a UAV-UGV docking and charging system in which an Unmanned Ground Vehicle (UGV) serves as a mobile charging station for a UAV. The docking station uses spring-loaded contacts to supply a 1C charging rate to the UAV, and employs a passive landing mechanism to enable reliable landing despite misalignment by the UAV. The UAV and UGV communicate using ESP32 microcontrollers, allowing the UAV to autonomously initiate and manage charging upon docking. Electromagnets secure the charging connection despite perturbations, and a vision-based guidance system using AprilTags and an OpenMV camera allows the UAV to land autonomously. We conducted 100 landing trials and achieved a 91\% success rate, validating the system's reliability. The design was further stress-tested on 45° inclines across multiple directions, confirming its robustness. This lightweight, adaptable system demonstrates a promising step toward resilient, infrastructure-free UAV operations and sets the foundation for future field-deployable UAV-UGV cooperative missions