MSE Research Project Database

Differential Privacy in Statistical Machine Learning


Project Leader: Ben Rubinstein
Student: Lingjuan Lyu,Maryam Fanaeepour,Leyla Roohi
Collaborators: Zuhe Zhang (Computing & Information Systems) Sanming Zhou (Mathematics & Statistics) Chris Culnane (Computing & Information Systems) Vanessa Teague (Computing & Information Systems), Francesco Alda (Bochum) Ben Fish (University of Illinois Chicago) Lev Reyzin (University of Illinois Chicago) Justin Bedo (WEHI)
Sponsors: Australian Research Council, National Science Foundation, Transport for NSW
Primary Contact: Ben Rubinstein (benjamin.rubinstein@unimelb.edu.au)
Keywords: artificial intelligence; computer security; machine learning
Disciplines: Computing and Information Systems
Domains: Networks and data in society

Machine learning research actively seeks data analyses that enjoy statistical efficiency (accurate predictions) and computational efficiency (scalability to largest datasets). Recently training data privacy has emerged as another dimension of performance for machine learning algorithms. This project explores the frontiers of achieving privacy and utility for machine learning through the formal framework of differential privacy. First proposed in 2006, the framework has become the leading approach to data privacy, has been awarded the 2017 Godel Prize for theoretical computer science, and has been deployed by Apple, Google and the US Census Bureau.

Further information: http://bipr.net

R package for differential privacy
test
Privately approximating nonprivate learning
test