Do Gradient Inversion Attacks Make Federated Learning Unsafe?
Do Gradient Inversion Attacks Make Federated Learning Unsafe?
Federated learning (FL) allows the collaborative training of AI models without needing to share raw data. This capability makes it especially interesting for healthcare applications where patient and data privacy is of utmost concern. However, recent works on the inversion of deep neural networks from model gradients raised concerns about …