EoN

This package is built based on the paper "Echo of Neighbors: Privacy Amplification for Personalized Private Federated Learning with Shuffle Model".

1. Package privAmp compare several privacy amplification results with shuffle model.
2. Package perS implement the framework APES and S-APES.

Usage

Privacy amplification in privAmp:

python evaluate_bounds.py

Framework APES and S-APES in perS:

python main.py --optimizer=personalLdpPgd --eps_dist=uniform --per_left=0.1 --per_right=2