11/20/2020: We are developing a new framework for backdoors with FL: Backdoors101. It extends to many new attacks (clean-label, physical backdoors, etc) and has improved user experience. Check it out!
backdoor_federated_learning
This code includes experiments for paper "How to Backdoor Federated Learning" (https://arxiv.org/abs/1807.00459)
All experiments are done using Python 3.7 and PyTorch 1.0.
mkdir saved_models
python training.py --params utils/params.yaml
I encourage to contact me ([email protected]) or raise Issues in GitHub, so I can provide more details and fix bugs.
Most of the experiments resulted by tweaking parameters in utils/params.yaml (for images) and utils/words.yaml (for text), you can play with them yourself.