this project is a reproduction code for Distributed Deep Learning, concening simple attack like Label Flipping and Bit Flipping (from this paper), and defence aggregation method Krum.
please download the CIFAR10 dataset in directory ./data/cifao10/
,
or automatically download when running code
Before attempting any further experiments, please confirm whether the configuration is correct in ./src/Config.py
if you'd like to test Lable Flipping attack, run:
python src/main.py --action 1
The result will be output in the console and saved in ./log/
and model will be saved in ./checkpoint
If you need to test other experiments, please check the action
parameter in ./src/main.py