This repo provides a federated learning framework that allows to carry out backdoor attacks under varying conditions. This is a keras-tensorflow implementation provided as a Jupyter Notebook which is hopefully self-explanatory.
Paper: Anomaly Localization in Model Gradients Under Backdoor Attacks Against Federated Learning https://arxiv.org/abs/2111.14683