Implementation of GANs Using Federated Learning
Raghda Ghonima · 2021
As the field of Generative Adversarial Networks (GANs) is flourishing so does security and privacy field. Unfortunately, both fields oppose each other’s progress. Since the standard GAN approaches require centralizing the training data into a common store so users’ data need to be transferred from users’ devices to the server opposing the basic concepts of data’s privacy and security. Users’ privacy and security were the primary motivation behind our paper. Furthermore, the architecture used allows for smarter models and less power consumption, all while ensuring privacy.