On the Protection of Private Information in Machine Learning Systems: Two Recent Approches

摘要

The recent, remarkable growth of machine learning has led to intense interestin the privacy of the data on which machine learning relies, and to newtechniques for preserving privacy. However, older ideas about privacy may wellremain valid and useful. This note reviews two recent works on privacy in thelight of the wisdom of some of the early literature, in particular theprinciples distilled by Saltzer and Schroeder in the 1970s.