This repository contains codes for paper Mutual Information Regularization for Vertical Federated Learning.
For the multi-classification task main task evalution, please use python vfl_main_task_no_defense.py while for Batch-level Label Inference Attack, please use python vfl_dlg_no_defense.py.
For binary classification please use python marvell_main.py --exp_type main_task or python marvell_main.py --exp_type attack for main task evaluation and direction scoring attack evaluation separately.
Config parameters can be specified like python vfl_dlg_no_defense.py --dataset cifar10.