CMU Team Wins PETs Prize

Adam KohlhaasThursday, April 6, 2023

A team of SCS professors and students won a PETS Prize for developing a framework to improve privacy-utility trade-offs in federated learning and applying that work to pandemic response and forecasting.

As personal and private information are increasingly digitized and shared, privacy-enhancing technologies (PETs) have become fundamental for protecting an individual's privacy while still allowing for the benefits of modern technology and data analysis. A team of Carnegie Mellon University researchers recently won the PETs Prize Challenge for their work to preserve privacy during pandemic forecasting. 

The United States and United Kingdom announced the PETs Prize Challenge in 2021 with the goal of using data to help tackle pressing global issues while upholding the essential right to privacy. These challenges centered on developing solutions to two specific problems: forecasting pandemic infection and detecting financial crime.

An interdisciplinary team of School of Computer Science professors and students developed a framework to improve privacy-utility trade-offs in federated learning, a technique that considers performing machine learning across private data silos. They applied their work to the challenges of pandemic response and forecasting, bringing home a first-place prize of $100,000. 

"In developing a solution to the pandemic forecasting problem, a key insight we had was that models of infection risk may differ slightly from one data silo (e.g., hospital, health district) to another," said Virginia Smith, an assistant professor in the Machine Learning Department (MLD). "Our team has spent several years researching techniques for personalized federated learning, which learn silo-specific models while simultaneously preserving data privacy. This challenge was a great way to put our research to the test to solve an important real-world problem."

In addition to Smith, the team included Ken Liu, a master's student in the Robotics Institute; Shengyuan Hu, a Ph.D. student in MLD; Tian Li, a Ph.D. student in the Computer Science Department; and Steven Wu, an assistant professor in the Software and Societal Systems Department.

For more information about the PETs Prize Challenge, visit the DrivenData website or read the White House press release from the Summit for Democracy.

For More Information

Aaron Aupperlee | 412-268-9068 |