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

   Abstract