National University of Singapore
As machine learning algorithms have become an influential component of critical decision making processes, the major question is whether they are trustworthy. Can we trust ML algorithms to perform accurately in noisy and adversarial environments, and be robust against adversarial data? Can we trust ML systems to have access to sensitive data during training and inference? Are there specific privacy risks of using machine learning models? Can we establish trust in black-box models, by providing interpretable predictions and explaining their decisions? Is it ethical to make use of machine learning algorithms? Can we trust them to be fair with respect to different individuals and groups? In this talk, I will discuss these fundamental concerns, and present various technical problems in this domain.
Reza Shokri is a NUS Presidential Young Professor of Computer Science. His research focuses on trustworthy machine learning, quantitative analysis of data privacy, and design of privacy-preserving algorithms for practical applications, ranging from data synthesis to federated learning. He is a recipient of the NUS Early Career Research Award 2019 for working on trustworthy machine learning. He received the Caspar Bowden Award for Outstanding Research in Privacy Enhancing Technologies in 2018, for his work on analyzing the privacy risks of machine learning models, and was a runner-up in 2012, for his work on quantifying location privacy. He obtained his PhD from EPFL.