Three-Body Problem in Privacy: A Disturbing Present and Promising Future by Dr Wenhai Sun

Privacy is a fundamental requirement for Internet users, and local differential privacy (LDP) has become a widely studied and deployed approach for protecting user data from service providers. Yet recent work has revealed a disturbing new reality: LDP systems themselves can be vulnerable to data poisoning attacks, blurring the traditional boundary between trusted and untrusted parties. In this setting, both users and service providers may become attackers or victims at the same time. This shift has major implications, not only for the practical deployment of privacy-preserving technologies, but also for the safety and freedom of people who rely on them most.

In this talk, Dr. Wenhai Sun will describe this emerging threat landscape and argue that it creates a new three-body problem for privacy research: balancing privacy, security, and utility in increasingly complex systems. He will present recent work on attacks and defenses in LDP-based data collection, and discuss how these challenges extend beyond classical data analytics to modern machine learning and AI systems. Finally, he will share opportunities for using machine intelligence itself to strengthen privacy protections, pointing toward a promising future of AI for privacy.