Inversion Influence Function

The paper discusses Deep Gradient Leakage (DGL), an attack that recovers private training images from gradient vectors, posing significant privacy challenges in distributed learning. To defend against such attacks, the…