Professor Xi He wins the Golden Jubilee Research Excellence Award

Professor Xi He has received the 2024 Faculty of Mathematics Golden Jubilee Research Excellence Award for the early-career category. 

Established in 2017, the Golden Jubilee Research Excellence Award recognizes faculty members for their outstanding research contributions. The selection is based on the nominee’s body of work or a single influential paper from the past five years.   

Every year, the University of Waterloo's Faculty of Mathematics honours one early-career and one mid-career faculty member. 

“Professor He is an influential academic researcher. Although she is one of our youngest faculty, joining us in 2019, she is poised to lead paradigm shifts in privacy and security research,” says Raouf Boutaba, University Professor and Director of the Cheriton School of Computer Science. “Congratulations, Professor He on receiving such a distinguished award!” 

The award, which is valued at $2,500, will be presented at the annual Fall Faculty Reception.  

“I am deeply honoured to receive this award. It truly motivates me to keep pushing the boundaries of data privacy research,” says Professor He. “I want to express my sincere gratitude to my family, colleagues, mentors, collaborators, and students for their unwavering support.” 

From 2008 to 2012, Professor He attended the National University of Singapore, graduating with a Bachelor of Science in Computer Science and Applied Mathematics. She completed her PhD in Computer Science at Duke University in 2018. 

Professor Xi He’s main research focus is on differential privacy, a rigorous mathematical framework. This technique allows users to collect and share information about a dataset, without compromising each data subject’s privacy. Namely, the user would inject controlled noise or randomness into the query results, which masks each contribution without impeding its accuracy. As a result, a user can effectively analyze a dataset while preventing any identification attacks or confidential leaks. 

With differential privacy, a business can collect and analyze user behaviour while controlling what is visible to their internal audiences like analysts. Similarly, government agencies can publish demographics and statistics about survey respondents while protecting their confidentiality. With the rise of technology, digital privacy is becoming a pressing societal concern. It has inspired new legislations like Canada’s Bill C-27 and Personal Information Protection and Electronic Documents Act, and the European Union’s General Data Protection Regulation. 

For her dissertation, Professor He proposed Blowfish Privacy, a novel differential privacy framework. A major challenge in differential privacy is that the user cannot specify which values in a dataset are related to each other for the same level of privacy protection. They must treat each entity separately or collectively. With Blowfish Privacy, the correlation is customizable: users can select which values must be protected and which ones may have been released to any potential attackers. This innovative approach has made data sharing and analysis more efficient and accurate and has inspired new privacy definitions for complex entity relationships used by the U.S. Census. Overall, Professor He’s work can bridge the gap between technical and mathematically rigorous privacy and the business and legal sphere. 

Professor He’s work extends to privacy data exploration. She was part of the Testbed for IoT-based Privacy-Preserving PERvasive Spaces (TIPPERS), an international collaboration including researchers from the Naval Information Warfare Center and the University of California, Irvine. The team focused on enhancing Internet of Things (IoT), devices that automatically connect and exchange data with each other through sensors, networks, and software. Although IoT can aid military personnel, such as providing them safety and real-time situational awareness about their equipment like ships or drones, they are prone to information leaks. For example, data from motion sensors could reveal a soldier’s base arrival and departure times. Instead, TIPPERS incorporates various privacy-enhancing technologies like encryption and differential privacy, to secure sensor data. In 2019, TIPPERS was deployed for the US Navy’s Trident Warrior experiment and produced promising results.  

Professor He’s other projects have intersected with fields like healthcare. At the Vector Institute, she is leading machine learning research in security, privacy and fairness, and developing scalable systems and applications for Canada’s healthcare system. In 2019, she collaborated with researchers from Northwestern University and Duke University to create Shrinkwrap, a private data federation (PDF). PDF is a collection of autonomous databases that lists out data points in a single user interface. Unfortunately, it faces slowdown issues, as data holders must insert dummy tuples, a type of data value, to pad and protect the database. As a result, PDFs do not scale well to complex structured query language (SQL) queries over large datasets. Instead, Shrinkwrap leverages differential privacy principles to prevent information leakages while accelerating the query process. Some potential applications include a clinical data research network, which allows healthcare sites to share data for research purposes, while protecting sensitive data. Overall, Professor He’s research has benefitted various fields such as military operations and healthcare.  

She also explored machine learning and synthetic data generation.  Her research on hyperparameter tuning revealed the missing privacy cost often left out in the private machine learning literature. It also urged the design of end-to-end private and useful optimizers.  

Notably, she co-created Kamino, a data synthesis system that ensures differential privacy while preserving the original dataset’s structure and correlations. It can generate useful synthetic data for applications of training classification models and answering marginal queries — and may surpass other state-of-the-art methods. Kamino's main innovation weaves two distinct paradigms: the randomized nature of differential privacy and the deterministic nature of database constraints.  

Professor He is recognized internationally for her contributions and expertise. She was invited to present her work on differential privacy at the International Association of Privacy Professions— the world’s largest information privacy community. According to CSRankings, she is one of North America’s top researchers in database systems.

Professor He has received several accolades for her cutting-edge work. She received Google’s Data Analytics and Insights Award (2023), Meta’s Privacy-preserving Technology Research Award (2022) and Google’s Processing and Analysis Award (2022) for her research on developing practical tools and systems with privacy guarantees. In 2021, she was awarded the Canada Foundation for Innovation’s John R. Evans Leader Fund for her project on scalable end-to-end differentially private machine learning. Her other awards include Duke University’s Outstanding Dissertation Award (2017), Google’s Ph.D. Fellowship in Privacy and Security (2017), and the Best Demonstration Award at the International Conference on Very Large Data Bases (2016). Most notably, she was appointed a Canada CIFAR AI Chair at the Vector Institute (2022).