Inverting Gradients - How easy is it to break privacy in federated learning?

NeurIPS 2020