Welcome to my personal website. I am Hannah, a Postdoctoral Scholar at the EECS Department of UC Berkeley, working in the NetSys Lab. My research interests span various aspects of system and network performance, privacy, security, and Internet of Things (IoT) protocols. Currently, my research focuses on the privacy and security dimensions of 5G networks. Previously, I have explored data-driven approaches to wireless communications and networking. My doctoral dissertation specifically concentrated on designing Media Access Control (MAC) layer protocols for WLANs, aligned with the recent IEEE 802.11 standards (11ac/ax, and upcoming be). For an in-depth view of my past and present projects, please refer to the Projects page. In addition to my role at UC Berkeley, I hold positions as a Visiting Research Scientist at the LIG Lab, supervised by Prof. Franck Rousseau; Research Associate at Inria Lyon; and Visiting Professor at North Carolina A&T. I am also collaborating with CISPA research institute (Germany), working alongside Dr. Singh. In the past, I worked as a Senior Wireless Engineer at Qualcomm, a Research Intern at Skylark Wireless, and a Research Assistant at Columbia University, under the guidance of Prof. Gil Zussman, and several more.
I'm Founding Director of Planet-X and our mission there is to represent an avant-garde concept where emerging technologies like AI, IoT, immersive reality, and advanced wireless technologies including 5G, 6G, NextG, mm-wave, and terahertz converge to create a world that is not only fast and connected but also agile, intelligent, secure, and sustainable.
Faros® is a true Massive MIMO wireless solution that scales to any size deployment on any sub-6 GHz spectrum.
PHTYON 100%
NETWORKING 90%
ALGORITHMS 80%
DATA-DRIVEN 90%
MACHINE LEARNING 75%
VR-Aware MU-MIMO Optimization
In this project, we propose a cross-layer optimization framework to enable mu-VR over 802.11ac/ax, a multi-user VR system for untethered mobile devices over 802.11ac/ax Wi-Fi. Our cross-layer design introduces novel optimizations in both application and wireless lower layers. Multi-user MIMO (MU-MIMO) is a technique in 802.11ac and 802.11ax that improves spectral efficiency by allowing concurrent communication between one AP and multiple clients. In practice, the expected gain is not always achieved and is sometimes even negative. 802.11ac/ax joint MU-MIMO user grouping and scheduling is crucial for multi-user applications. Although Mu-MIMO is introduced in 802.11ac and ax to improve spectral efficiency by allowing concurrent communication, it may introduce high delays and low throughput if AP selects the wrong users to group in a MU-MIMO transmission, as users with correlated channels cause high packet losses due to interference. Therefore, using a commodity 802.11ac AP, we first experimentally show that factors such as user mobility or user device type are important to both MU-MIMO MAC and application layer optimizations. Based on this observation, the AP can have a predictive approach to decide whether a user can benefit from participating in MU-MIMO rather than current reactive algorithms. We present our design and its evaluation on COTS smartphones and laptops over 802.11ac Wi-Fi.
Data-Driven MAC Protocol Design Optimization
Networking protocols are designed through long-time and hard-work human efforts. Machine Learning (ML)-based solutions have been developed for communication protocol design to avoid manual efforts to tune individual protocol parameters. While other proposed ML-based methods mainly focus on tuning individual protocol parameters (e.g., adjusting contention window), our main contribution is to propose a novel Deep Reinforcement Learning (DRL)-based framework to systematically design and evaluate networking protocols. We decouple a protocol into a set of parametric modules, each representing a main protocol functionality that is used as DRL input to better understand the generated protocols design optimization and analyze them in a systematic fashion. As a case study, we introduce and evaluate DeepMAC a framework in which a MAC protocol is decoupled into a set of blocks across popular flavors of 802.11 WLANs (e.g., 802.11 a/b/g/n/ac). We are interested to see what blocks are selected by DeepMAC across different networking scenarios and whether DeepMAC is able to adapt to network dynamics.
Online Video Rate Adaptation
Video delivery over wireless networks is challenging due to the lack of spectrum and reliability issues. This challenge is exacerbated in dense venues. To address this issue, Wi-Fi multicast with the ability to simultaneously multicast the same video contents to a group of users, has gained attention. For successful video delivery, the content providers are interested in evaluating the performance of such traffic from the final users'' perspective, that is, their Quality of Experience (QoE). The QoE ties together user perception, experience, and expectations to application and network performance. However, ensuring high QoE for multicast video streaming is challenging. Although, there have been considerable efforts in the literature to design Adaptive Bitrate (ABR) streaming algorithms to ensure the video QoE, applying these approaches to wireless multicast is not straightforward due to lack of feedback and unreliable transmissions. To overcome these issues, transmission and video rate can be jointly controlled to ensure the video QoE using Wi-Fi multicast. In this project, we have collaborated with the Adaptive Multicast Services (AMuSe) project at wim.net Lab at Columbia University which is an end to end system for high quality video delivery to a large number of users in dense environments which leverages Wi-Fi multicast. The project involves improvements of the DYnamic Video and Rate (DYVR) algorithm which is an online control algorithm for jointly multicast transmission and video rates adaptation. We present a new channel estimation method for this algorithm. We evaluate the algorithm using the new channel estimation through extensive experiments in a push-based platform consisting of Android devices and an off-the-shelf wireless Access Point (AP). We show that our new channel estimation method improves the total performance of the algorithm significantly. We also compare two distinct versions of this algorithm against current ABR based state-of-the-art approaches.
Collaborative Video Analytics at the Edge
Today, video cameras are deployed in dense for monitoring physical places e.g., city, industrial, or agricultural sites. In the current systems, each camera node sends its feed to a cloud server individually. However, this approach suffers from several hurdles including higher computation cost, large bandwidth requirement for analyzing the enormous data, and privacy concerns. In dense deployment, video nodes typically demonstrate a significant spatio-temporalcorrelation. To overcome these obstacles in current approaches, this project introduces CONVINCE, a new approach to look at the network cameras as a collective entity that enables collaborative video analytics pipeline among cameras. CONVINCE aims at 1) reducing the computation cost and bandwidth requirements by leveraging spatio-temporal correlations among cameras in eliminating redundant frames intelligently, and ii) improving vision algorithms'' accuracy by enabling collaborative knowledge sharing among relevant cameras. Our results demonstrate that CONVINCE achieves an object identification accuracy of ?91\\\%, by transmitting only about ?25\\\% of all the recorded frames.