Vector Researcher Xi He Advances Differential Privacy for Data Protection
Xi He, a Vector Institute Faculty Member and professor at the University of Waterloo, is pioneering advancements in differential privacy to safeguard personal data in artificial intelligence systems. Addressing concerns about surveillance capitalism and data misuse, He emphasizes the critical balance between data utility and individual privacy. Her research introduces a framework that adds controlled noise to datasets, ensuring that analysts cannot determine if specific individuals contributed to the results. Unlike traditional models that maximize utility under privacy constraints, He’s approach optimizes algorithms to achieve specific accuracy goals with minimum privacy cost. For instance, in retail scenarios, this method obscures whether a single customer purchased an item while still providing useful aggregate trends. He aims to generalize these systems into scalable, plug-and-play database frameworks that allow organizations to process sensitive data securely. As a Canada CIFAR AI Chair, she seeks to make trustworthy, privacy-preserving AI accessible beyond limited applications by major tech companies, ensuring personal information remains protected from unauthorized access or discriminatory use in sectors like insurance and finance.
Wire timeline
Vector Researcher Xi He Advances Differential Privacy for Data Protection
Xi He, a Vector Institute Faculty Member and professor at the University of Waterloo, is pioneering advancements in differential privacy to safeguard personal data in artificial intelligence systems. Addressing concerns about surveillance capitalism and data misuse, He emphasizes the critical balance between data utility and individual privacy. Her research introduces a framework that adds controlled noise to datasets, ensuring that analysts cannot determine if specific individuals contributed to the results. Unlike traditional models that maximize utility under privacy constraints, He’s approach optimizes algorithms to achieve specific accuracy goals with minimum privacy cost. For instance, in retail scenarios, this method obscures whether a single customer purchased an item while still providing useful aggregate trends. He aims to generalize these systems into scalable, plug-and-play database frameworks that allow organizations to process sensitive data securely. As a Canada CIFAR AI Chair, she seeks to make trustworthy, privacy-preserving AI accessible beyond limited applications by major tech companies, ensuring personal information remains protected from unauthorized access or discriminatory use in sectors like insurance and finance.
Vector Institute for Artificial Intelligence