AI-secure / GMI-Attack

24 stars 10 forks source link

GMI-Attack

This repo contains code for paper "The Secret Revealer: Generative Model-Inversion Attacks Against Deep Neural

Networks"

Dependencies

numpy>=1.16.3

pytorch>=1.2.0

Usage

train_gan.py: Train the attack models including generator network and discriminator network with public data

attack.py: Implement model inversion attack on target models which is trained on private data

Celeba/masked/train_inv.py: Train the attack models with auxiliary knowledge