Differential Privacy Preservation in Deep Learning
Dr. Hai Phan, New Jersey Institute of Technology
Friday, February 24, 2017
9:30am - 10:30am
In recent years, advances in deep learning have enabled a dizzying array of applications such as data analytics, signal and information processing, and autonomous systems. For instance, deep learning has applications in a number of healthcare areas, e.g., phenotype extraction and health risk prediction, prediction of the development of various diseases including schizophrenia, a variety of cancers, diabetes, heart failure, and many more. This presents an obvious threat to privacy in new deep learning systems and models, which are being developed and deployed. In this talk, I will review the current picture of preserving privacy in deep learning. Then, I will introduce our approaches in differential privacy preservation in deep learning. The key characteristics of our frameworks are: (1) It is totally independent of the data size, and (2) It has the ability to redistribute the noise injected into deep neural networks towards an improvement of usability. Future directions of our research will be briefly introduced as well.
Hai Phan is an Assistant Professor at the Ying Wu College of Computing, New Jersey Institute of Technology. He holds a Ph.D. in computer science, awarded in 2013 by the French National Centre for Scientific Research (CNRS) - University Montpellier 2 (UM2). His topics of interest primarily focus on data science, machine learning, deep learning, and privacy and security, especially for health informatics, social network analysis, and spatio-temporal data mining. His contributions have been published in at least 25 publications at leading and top-tier venues, such as AAAI, ACM Multimedia, ACM CIKM, ACM TIST, IEEE Intelligent Systems, KAIS, SDM, ECML-PKDD, ACM GIS, etc. Several of them had been selected as best papers and two contributions had been presented as invited talks.