A Survey on Privacy Preserving Deep Learning System

PhD Qualifying Examination


Title: "A Survey on Privacy Preserving Deep Learning System"

by

Mr. Chaoliang ZENG


Abstract:

Over the past years, deep learning has achieved tremendous successes in 
multiple application domains. A human-comparable neural network model 
comes from expert-elaborate model design, massive and high-quality data 
feed, and great computing power support. However, given the increasing 
attention over privacy, the collaboration to train a model becomes 
challenging.

To solve the dilemma, several privacy-preserving deep learning systems are 
designed. In this survey, we first give an overview of neural networks and 
challenges on privacy- preserving deep learning. Then we discuss the basic 
security methods, including differential privacy, homomorphic encryption, 
secure multi-party computation, and trusted execution environment., that 
can be adopted in privacy-preserving deep learning. We also introduce the 
corresponding system designs based on these basic methods. Last, we take 
one step further and introduce optimizations for utility, privacy, and 
performance from two directions, i.e., working environment and neural 
network layer properties.


Date:			Thursday, 2 July 2020

Time:                  	4:00pm - 6:00pm

Zoom meeting:           https://hkust.zoom.us/j/92019653963

Committee Members:	Dr. Kai Chen (Supervisor)
 			Dr. Yangqiu Song (Chairperson)
 			Dr. Dimitrios Papadopoulos
 			Dr. Wei Wang


**** ALL are Welcome ****