Research & Publications
6+ years of research experience in machine learning, networking, anomaly detection, and edge/cloud computing.
5 Journal Articles
6 Conference Papers
1 US Patent
Publications
Research & Development Projects
Multi-Type Water Contaminant Identification Using IoT Sensors
12/2022 – 10/2023- •Developed a contaminant identification system that combines Electrochemical Impedance Spectroscopy (EIS) with the Internet of Things and Machine Learning techniques.
- •Achieve low cost and scalability at network edge while offering unprecedented identification sensitivity and selectivity for public health.
AI-based Smart Home Devices Privacy Inference and Defense
08/2022 – 05/2023- •Explored vulnerabilities of IoT wireless network traffic leaking user activities and motion traces at home.
- •Developed a machine learning (random forest, DBSCAN, autoencoder)-based IoT device event identification solution using encrypted wireless traffic data without access to home network.
- •Performed packet padding, traffic injection, and traffic shaping techniques to reduce the feasibility of inference attacks.
Verifiable Edge-aided IoT Computation Outsourcing
09/2021 – 07/2022- •Designed lightweight and collusion-resistant verification protocols executing on resource-constrained IoT devices.
- •Proposed algorithm to masquerade verification from regular device computation requests.
- •Implemented protocols on NS-3 network simulator (C++) to validate 8x less overhead and high robustness.
Intelligent Resource Allocation for Heterogeneous Edge-IoT System
09/2020 – 04/2021- •Modeled the patterns of computing requests from heterogeneous IoT devices with Semi-Supervised machine learning models (One Class-SVM, One Class Neural Network) using scikit-learn and Keras.
- •Proposed a rational and fair resource allocation model by formulating a novel Markov Decision Process that allocates computing resources of 100 edge servers to 500 IoT devices.
Context-aware IoT Device Activity Anomaly Detection
05/2019 – 04/2020- •Collected and analyzed network traffic data of heterogeneous IoT devices using tcpdump and WireShark tools.
- •Extracted relevant features characterizing network behavior using Mutual Information analysis.
- •Designed a Conditional Variational Autoencoder model with Keras/TensorFlow framework achieving 98% anomaly detection accuracy.
Blockchain-based Secure Resource Management for Edge-IoT System
08/2017 – 06/2018- •Designed a blockchain-based registration and management system of IoT devices tested on both Ethereum and Hyperledger platforms.
- •Developed smart contracts using Solidity programming language to enforce access control rules of edge computing requests.
- •Implemented a web interface to register and query IoT management information using jQuery and Node.js.
Robust Intrusion Detection Against Adversarial Learning Attacks
07/2018 – 04/2019- •Investigated vulnerabilities of Deep Neural Networks on adversarial attacks causing misclassification.
- •Developed an ensemble defense mechanism using generative adversarial networks and adversarial learning techniques against four state-of-the-art attacks with 97% success rate.
Edge-cloud Computing Testbed on University Campus
09/2016 – 04/2017- •Deployed an OpenStack-based edge computing platform using Dell rack servers on the campus network.
- •Provided lab-use cloud-based virtual machines for instructors and students.