Learning From Distributed Private Data: Algorithms and Applications

Distributed learning from biomedical data is often hindered by ethical, legal, and technological concerns about data sharing. Data holders wish to maintain control over the uses of their data, and patients or study subjects may be hesitant to allow free and open use of their private medical data. Differential privacy is a framework which allows us to quantify privacy risk. In privacy-preserving distributed learning, the data stays at each site: they locally compute a privacy-preserving summary of their information. The summaries are sent to a private aggregator that performs the final analysis. Differentially private algorithms guarantee privacy by deliberately introducing some noise into the computation "“ the uncertainty from the noise masks individual data points. This leads to a tradeoff between privacy and accuracy. In this talk I will discuss algorithms for privacy-preserving learning as well as a recent proof-of-concept for this approach applied to neuroimaging data for mental health research.
Anand D. Sarwate joined as an Assistant Professor in the Department of Electrical and Computer Engineering at Rutgers University in 2014. He received B.S. degrees in Electrical Engineering and Mathematics from MIT in 2002, an M.S. in Electrical Engineering from UC Berkeley in 2005 and a PhD in Electrical Engineering from UC Berkeley in 2008. From 2008-2011 he was a postdoctoral researcher at the Information Theory and Applications Center at UC San Diego and from 2011-2013 he was a Research Assistant Professor at the Toyota Technological Institute at Chicago. He is the recipient of the NSF CAREER award in 2015. His research areas are in information theory, machine learning, and signal processing, with focus on distributed inference and learning, privacy and security, and applications in biomedical research.