Yuan Tian
When AI Meets Privacy: Building Privacy-preserving AI Systems
Thursday, February 17th, 2022 @ 2:00 p.m. CST
This talk has ended.
A recording of the talk will be posted for CS faculty and students to view within 24 hours.
Computing is undergoing a significant shift. First, the explosive growth of the Internet of Things (IoT) enables users to interact with computing systems and physical environments in novel ways through perceptual interfaces (e.g., microphones and cameras). Second, machine learning algorithms collect huge amounts of data and make critical decisions on new computing systems. While these trends bring unprecedented functionality, they also drastically increase the number of untrusted algorithms, implementations, interfaces, and the amount of private data processed by them, endangering user privacy. The pressing question is how to protect user privacy with utility/performance preserved in machine learning? The challenges are two folds: (1) how to improve user privacy in light of the rapid advancement of AI; (2) how to prevent the existing AI from leaking private user data.
In this talk, I’ll introduce my work on protecting user privacy in machine learning systems from practical applications to theoretical frameworks. First, I will use conversational AI systems to show our research in identifying privacy violations with machine learning. Second, I will talk about our research for building scalable and accurate privacy-preserving machine learning systems. I will present our work CryptGPU, the first privacy-preserving secure Multi-Party Computation framework fully implemented on the GPU, which scales to modern large models and datasets.