Poisoners for label flipping and backdooring. It is now ready for label flipping with contribution measurements.
- py3.7 + pip install -r requirements.pip
- generate_data_distribution.py
- generate_default_models.py
- label_flipping_attack.py
- The arguments are in 'arguments.py' + label_flipping_attack.py.
- The traning round (epoch in this repository) should be set to 200.
5201 5202 5203: fashion-maverick-1-fedprox-mu0.01 5204 5205 5206: cifar-maverick-1-fedprox-mu0.01