A decentralized and private IoT Framework using Blockchain
While billions of IoT devices—from smart home appliances to wearable trackers—collect and share data seamlessly,
they expose sensitive user information to severe privacy risks. Because resource-constrained IoT devices lack
the computational power and memory to run traditional security algorithms, a lightweight alternative is essential.
This project aims to design and implement a decentralized, privacy-preserving IoT framework using blockchain technology.
Our research focuses on securing data sharing and access control without relying on centralized servers or heavy computational overhead.
Wireless Networks
Barrier Coverage on WSNs and VANETs
Connected vehicles in Vehicular Ad-hoc Networks (VANETs), equipped with wireless transceivers and cameras,
offer new possibilities for intelligent urban monitoring. The goal of this research is to construct
an impromptu surveillance system based on barrier coverage using Wireless Sensor Networks (WSNs) and VANET nodes.
By surrounding a specific area of interest, such as a city block, our framework detects and tracks targets attempting
to cross the perimeter on foot or by vehicle. This work provides a rapid, infrastructure-free emergency response solution
for public safety and security scenarios.
Network Performance Evaluation on Interference of Things
High node mobility in dense wireless environments frequently causes co-channel and adjacent-channel interference,
leading to severe network performance degradation. This project aims to analyze cross-layer error propagation under
wireless interference and validate the accuracy of NS-3 interference models. Through both outdoor field experiments
and NS-3 simulations, we investigate how physical-layer interference affects upper network layers up to TCP.
Our findings provide practical insights for improving wireless ad-hoc network reliability and simulation fidelity.
Machine Learning
Rapid Thrombogenesis Prediction in Covid-19 patients using Machine Learning
Traditional computational fluid dynamics (CFD) methods for evaluating blood clot risks in patient-specific
anatomies require weeks to months of computation per patient. To overcome this bottleneck, this project develops
a machine learning-based framework for the rapid prediction of thrombogenesis in the carotid artery of COVID-19 patients.
By training ML models on hemodynamic metrics, our system swiftly assesses thrombosis risk in real time,
supporting early diagnosis and preventive medical care.