CoPur: Certifiably Robust Collaborative Inference via Feature Purification

This repository is the official implementation of "CoPur: Certifiably Robust Collaborative Inference via Feature Purification".

Download and Installation

The required packages can be installed by:

pip install -r requirement.txt

For datasets:

• Download NUS_WIDE dataset to folder nus-wide/NUS_WIDE
• The Extrasensory dataset is available in extrasensory/example.csv

Usage

Training and testing on NUS-WIDE dataset:

0. Enter the folder

cd nus-wide

1. The pre-trained Local feature extractors can be found in the folder nus_results
2. Feature subspace learning

python nus_ae_train.py --ae_epochs 400 --ae_lr 0.001 --img_feature_div 0 360 634 --text_feature_div 0 500 1000

3. Fusion center training

python nus_server_ae_train.py --use_ae --lr 0.001 --epochs 200 --img_feature_div 0 360 634 --text_feature_div 0 500 1000

4. Inference under untargeted attack

python nus_infer_attack.py --use_ae --L_lr 0.001 --img_feature_div 0 360 634 --text_feature_div 0 500 1000 --corruption_amp 10

5. Inference under targeted attack

python nus_infer_attack.py --use_ae --L_lr 0.001 --img_feature_div 0 360 634 --text_feature_div 0 500 1000 --num_test_samples 1000 --targeted 1

Training and testing on Extrasensory dataset:

0. Enter the folder

cd extrasensory

1. Local feature extractor training (trained models are saved in `nus_results' folder)

python sensor_normal_train.py --epochs 200 --emb_dim 32 --num_class 1 --lr 0.001 --text_feature_div 0 26 52 83 129 138 155 183 209 213 221

2. Feature subspace learning

python sensor_ae_train.py --ae_epochs 400 --ae_lr 0.001 --text_feature_div 0 26 52 83 129 138 155 183 209 213 221

3. Fusion center training

python sensor_server_ae_train.py --use_ae --epochs 200 --lr 0.001 --text_feature_div 0 26 52 83 129 138 155 183 209 213 221

4. Inference under untargeted attack

python sensor_infer_attack.py --use_ae  --L_lr 0.001 --text_feature_div 0 26 52 83 129 138 155 183 209 213 221 --corruption_amp 10 

5. Inference under targeted attack

python sensor_infer_attack.py --use_ae  --L_lr 0.001 --text_feature_div 0 26 52 83 129 138 155 183 209 213 221 --targeted 1 

Citation

If you find our work useful in your research, please consider citing:

@inproceedings{
liu2022copur,
title={CoPur: Certifiably Robust Collaborative Inference via Feature Purification},
author={Jing Liu and Chulin Xie and Oluwasanmi O Koyejo and Bo Li},
booktitle={Advances in Neural Information Processing Systems},
editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},
year={2022},
url={https://openreview.net/forum?id=r5rzV51GZx}
}