Have your data and share it too: Private data analytics at scale

As the importance of data protection and privacy legislation is increasingly recognized worldwide, protecting sensitive information and individual privacy presents a major challenge for modern big data analytics systems. The ever-growing list of major data breaches (and associated fines) clearly demonstrates the inadequacy of earlier ad-hoc solutions to the problem, as well as the need to effectively bridge legal and technical/systems interpretations of data privacy.As the importance of data protection and privacy legislation is increasingly recognized worldwide, protecting sensitive information and individual privacy presents a major challenge for modern big data analytics systems. The ever-growing list of major data breaches (and associated fines) clearly demonstrates the inadequacy of earlier ad-hoc solutions to the problem, as well as the need to effectively bridge legal and technical/systems interpretations of data privacy. In this talk, I will present different modern privacy enhancing technologies (including federated learning, secure computing, differential privacy, and synthetic data), and discuss how they can enable formal, cryptographic notions of privacy in large-scale data analytics. The focus will be on our recent efforts to build systems and tools to support querying and machine learning over sensitive medical data. Several open challenges and directions for future research will also be discussed.