Hyperlocal Risk Monitoring and Pandemic Preparedness through Privacy-Enhanced Mobility and Social Interactions Analysis
Hyperlocal risk monitoring is critical for gaining a better estimate of the current and future infection risk at a population level during a pandemic, as well as a better understanding of disparate infection risk in vulnerable groups. However, many challenges remain in enabling hyperlocal risk monitoring and decision making for community-based pandemic preparedness. First, most popular disease prediction models are at a coarse-grained level without considering mobility and social interactions data. Second, a one-size-fits-all approach fails to appropriately address heterogeneity in mobility patterns and interactions, which can be highly community or country specific (e.g., US vs. Japan) and the risks are affected by regional, socioeconomic, behavioral, and cultural differences. Finally, privacy concerns limit the access and use of fine-grained mobility and social interactions data. This project represents a multi-disciplinary collaboration between US and Japanese researchers including lead institutions at Emory University and Kyoto University. The project includes strong community engagement with communities in the US (primarily Georgia and Southern California) and Japan (primarily Kyoto prefecture) as well as local and regional health centers.. The project aims to develop a framework for privacy-enhanced monitoring and analysis of fine-grained mobility and social interactions data to enable hyperlocal risk monitoring and data-driven decision-making. Such hyperlocal situational awareness can help governments and response officials at all levels (from schools and businesses to county and state) for policy making, e.g., open in-person or online; close or partially shut down; and reallocate medical supplies and workforces to vulnerable areas. It can also benefit an individual's personal decision making in the community, e.g., to avoid high-risk areas.
The project includes an integrative research agenda that addresses both technical and social science questions to enable hyperlocal data collection, analysis, and decision making: 1) develop computational and modeling methods for fine-grained risk estimation and scenario analysis (e.g., future estimated risk under partial shutdown) by incorporating real-world mobility and social interactions data; 2) study how mobility patterns, social interactions, behaviors, and risks change and differ by region, socioeconomic status, and country using the US vs. Japan as exemplars; and 3) develop privacy-enhancing technologies and study their social adoption and legal implications for collection and aggregation of mobility and social network data. The team will engage with community stakeholders across the entire data-driven decision-making pipeline including data providers; local public health agencies; local decision makers; and community members. The goal is to not only build a data aggregation and analytics platform but also a feedback loop that enables data-driven policy and decision making while simultaneously enabling social scientists, epidemiologists, and decision makers to steer the data collection, aggregation, and analysis, ultimately enabling better preparation and readiness for future outbreaks.
This project is a joint collaboration between the National Science Foundation and the Japan Science and Technology Agency
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
-
Performance PeriodOctober 2021 - September 2024
-
Emory University
-
Award Number2125530
-
Lead PILi Xiong
-
Co-PIWeihua An
-
Co-PIShivani Patel
-
Co-PICyrus Shahabi
Project Material
- Echo of Neighbors: Privacy Amplification for Personalized Private Federated Learning with Shuffle Model
- RobustFed: A Truth Inference Approach for Robust Federated Learning
- Federated Pruning: Improving Neural Network Efficiency with Federated Learning
- PrivLBS: Local Differential Privacy for Location-Based Services with Staircase Randomized Response
- Communication Efficient Tensor Factorization for Decentralized Healthcare Networks
- Federated graph classification over non-iid graphs
- Projected federated averaging with heterogeneous differential privacy
- A population-based study of the trend in SARS-CoV-2 diagnostic modalities from the beginning of the pandemic to the Omicron surge in Kyoto City, Kyoto, Japan
- Supporting secure dynamic alert zones using searchable encryption and graph embedding
- IGAMT: Privacy-Preserving Electronic Health Record Synthesization with Heterogeneity and Irregularity
- Racial Disparities in COVID-19 Severity Are Partially Mediated by Chronic Stress—Evidence from a Large Integrated Healthcare System
- Supporting Pandemic Preparedness with Privacy Enhancing Technology
- Personalized Differentially Private Federated Learning without Exposing Privacy Budgets
- MUter: Machine Unlearning on Adversarial Training Models
- Closed-form Machine Unlearning for Matrix Factorization
- ShapleyFL: Robust Federated Learning Based on Shapley Value
- Equitable Data Valuation Meets the Right to Be Forgotten in Model Markets
- CSGAN: Modality-Aware Trajectory Generation via Clustering-based Sequence GAN
- NeuroSketch: Fast and Approximate Evaluation of Range Aggregate Queries with Neural Networks
- Efficient Sampling Approaches to Shapley Value Approximation
- A Neural Approach to Spatio-Temporal Data Release with User-Level Differential Privacy
- Federated Node Classification over Graphs with Latent Link-type Heterogeneity
- Dynamic Shapley Value Computation
- Models and mechanisms for spatial data fairness
- Toward Accurate Spatiotemporal COVID-19 Risk Scores Using High-Resolution Real-World Mobility Data
- HAGEN: Homophily-Aware Graph Convolutional Recurrent Network for Crime Forecasting
- Differentially Private Occupancy Monitoring from WiFi Access Points
- Differentially-Private Publication of Origin-Destination Matrices with Intermediate Stops
- A neural database for differentially private spatial range queries
- Estimating spread of contact-based contagions in a population through sub-sampling
- Social Robot Design Challenge: Gathering design requirements from teens
Li Xiong is a Samuel Candler Dobbs Professor of Computer Science and Professor of Biomedical Informatics at Emory University. She held a Winship Distinguished Research Professorship from 2015-2018. She has a Ph.D. from Georgia Institute of Technology, an MS from Johns Hopkins University, and a BS from the University of Science and Technology of China. Her research lab, Assured Information Management and Sharing (AIMS), conducts research in the intersection of data management, machine learning, and data privacy and security, with a recent focus on privacy-enhancing and trustworthy machine learning and data sharing algorithms to advance data driven-AI systems for healthcare, public health, and spatial intelligence. She has published over 180 papers and received six best paper or runner up awards. She has served and serves as associate editor for IEEE TKDE, IEEE TDSC, and VLDBJ, general chair for ACM SIGSPATIAL 2024, CIKM 2022, program chair for IEEE BigData 2020 and ACM SIGSPATIAL 2020, 2018, tutorial chair for VLDB 2024, program vice-chair for VLDB 2024, ACM SIGMOD 2024, 2022, and IEEE ICDE 2023, 2020. Her research is supported by federal agencies including NSF, NIH, IARPA, AFOSR, PCORI, and industry awards including Google, IBM, Mitsubishi, Cisco, AT&T, and Woodrow Wilson Foundation. She is an IEEE fellow.