A Systematic Review of Federated Generative Models
Ashkan Vedadi Gargary, Emiliano De Cristofaro
TL;DR
This systematic survey analyzes the intersection of Federated Learning and Generative Models from 2019 to 2024, synthesizing nearly 100 papers across GANs, VAEs, and diffusion models. It organizes methods into three strands: federated generation, attacks/defenses, and data heterogeneity mitigation, with structured review criteria covering model types, algorithms, data, DP usage, and code availability. A key finding is that Federated GANs dominate early work, with diffusion-based FL emerging as a strong alternative for convergence and communication efficiency, while one-shot FL and pre-trained diffusion/LLM-based approaches gain traction. The work highlights persistent challenges in scalability, cross-device settings, and tabular data, and provides a roadmap for future research in privacy-preserving, robust, and efficient federated generative modelling with practical impact for healthcare, finance, and IoT contexts.
Abstract
Federated Learning (FL) has emerged as a solution for distributed systems that allow clients to train models on their data and only share models instead of local data. Generative Models are designed to learn the distribution of a dataset and generate new data samples that are similar to the original data. Many prior works have tried proposing Federated Generative Models. Using Federated Learning and Generative Models together can be susceptible to attacks, and designing the optimal architecture remains challenging. This survey covers the growing interest in the intersection of FL and Generative Models by comprehensively reviewing research conducted from 2019 to 2024. We systematically compare nearly 100 papers, focusing on their FL and Generative Model methods and privacy considerations. To make this field more accessible to newcomers, we highlight the state-of-the-art advancements and identify unresolved challenges, offering insights for future research in this evolving field.