The official implementation of NeurIPS 2021 paper: Finding Optimal Tangent Points for Reducing Distortions of Hard-label Attacks

Overview

Introduction

This repository includes the source code for "Finding Optimal Tangent Points for Reducing Distortions of Hard-label Attacks", which is published in NeurIPS 2021.

Citation

We kindly ask anybody who uses this code to cite the following bibtex:

@inproceedings{
    ma2021finding,
    title={Finding Optimal Tangent Points for Reducing Distortions of Hard-label Attacks},
    author={Chen Ma and Xiangyu Guo and Li Chen and Jun-Hai Yong and Yisen Wang},
    booktitle={Thirty-Fifth Conference on Neural Information Processing Systems},
    year={2021},
    url={https://openreview.net/forum?id=g0wang64Zjd}
}

Structure of Folders and Files

+-- configures
|   |-- HSJA.json  # the hyperparameters setting of HSJA, which is also used in Tangent Attack
+-- dataset
|   |-- dataset_loader_maker.py  # it returns the data loader class that includes 1000 attacks images for the experiments.
|   |-- npz_dataset.py  # it is the dataset class that includes 1000 attacks images for the experiments.
+-- models
|   |-- defensive_model.py # the wrapper of defensive networks (e.g., AT, ComDefend, Feature Scatter), and it converts the input image's pixels to the range of 0 to 1 before feeding.
|   |-- standard_model.py # the wrapper of standard classification networks, and it converts the input image's pixels to the range of 0 to 1 before feeding.
+-- tangent_attack_hemisphere
|   |-- attack.py  # the main class for the attack.
|   |-- tangent_point_analytical_solution.py  # the class for computing the optimal tagent point of the hemisphere.
+-- tangent_attack_semiellipsoid
|   |-- attack.py  # the main class for the attack.
|   |-- tangent_point_analytical_solution.py  # the class for computing the optimal tagent point of the semi-ellipsoid.
+-- cifar_models   # this folder includes the target models of CIFAR-10, i.e., PyramidNet-272, GDAS, WRN-28, and WRN-40 networks.
|-- config.py   # the main configuration of Tangent Attack.
|-- logs  # all the output (logs and result stats files) are located inside this folder
|-- train_pytorch_model  # the pretrained weights of target models
|-- attacked_images  # the 1000 image data for evaluation 

In general, the train_pytorch_model includes the pretrained models' weights, and attacked_images includes the image data, which is packaged into .npz format with pixel range of [0-1].

In the attack, all logs are dumped to logs folder, the statistical results are also written into logs folder, which are .json format.

Attack Command

The following command could run Tangent Attack (TA) and Generalized Tangent Attack (G-TA) on the CIFAR-10 dataset under the untargetd attack's setting:

python tangent_attack_hemisphere/attack.py --gpu 0 --norm l2 --dataset CIFAR-10 --arch resnet-50
python tangent_attack_hemisphere/attack.py --gpu 0 --norm l2 --dataset CIFAR-10 --arch gdas
python tangent_attack_semiellipsoid/attack.py --gpu 0 --norm l2 --dataset CIFAR-10 --arch resnet-50
python tangent_attack_semiellipsoid/attack.py --gpu 0 --norm l2 --dataset CIFAR-10 --arch gdas

Once the attack is running, it directly writes the log into a newly created logs folder. After attacking, the statistical result are also dumped into the same folder, which is named as *.json file.

Also, you can use the following bash shell to run the attack of different models one by one.

./tangent_attack_CIFAR_undefended_models.sh

The commmand of attacks of defense models are presented in tangent_attack_CIFAR_defense_models.sh.

  • The gpu device could be specified by the --gpu device_id argument.
  • the targeted attack can be specified by the --targeted argument. If you want to perform untargeted attack, just don't pass it.
  • the attack of defense models uses --attack_defense --defense_model adv_train/jpeg/com_defend/TRADES argument.

Requirement

Our code is tested on the following environment (probably also works on other environments without many changes):

  • Ubuntu 18.04
  • Python 3.7.3
  • CUDA 11.1
  • CUDNN 8.0.4
  • PyTorch 1.7.1
  • torchvision 0.8.2
  • numpy 1.18.0
  • pretrainedmodels 0.7.4
  • bidict 0.18.0
  • advertorch 0.1.5
  • glog 0.3.1

You can just type pip install -r requirements.txt to install packages.

Download Files of Running Results and Logs

I have uploaded all the logs and results with the compressed zip file format onto this google drive link so that you can download them.

Owner
machen
machen
NALSM: Neuron-Astrocyte Liquid State Machine

NALSM: Neuron-Astrocyte Liquid State Machine This package is a Tensorflow implementation of the Neuron-Astrocyte Liquid State Machine (NALSM) that int

Computational Brain Lab 4 Nov 28, 2022
PiRapGenerator - Make anyone rap the digits of pi

PiRapGenerator Make anyone rap the digits of pi (sample files are of Ted Nivison

7 Oct 02, 2022
Learning to Reconstruct 3D Manhattan Wireframes from a Single Image

Learning to Reconstruct 3D Manhattan Wireframes From a Single Image This repository contains the PyTorch implementation of the paper: Yichao Zhou, Hao

Yichao Zhou 50 Dec 27, 2022
Validated, scalable, community developed variant calling, RNA-seq and small RNA analysis

Validated, scalable, community developed variant calling, RNA-seq and small RNA analysis. You write a high level configuration file specifying your in

Blue Collar Bioinformatics 917 Jan 03, 2023
Training, generation, and analysis code for Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics

Location-Aware Generative Adversarial Networks (LAGAN) for Physics Synthesis This repository contains all the code used in L. de Oliveira (@lukedeo),

Deep Learning for HEP 57 Oct 22, 2022
CIFAR-10 Photo Classification

Image-Classification CIFAR-10 Photo Classification CIFAR-10_Dataset_Classfication CIFAR-10 Photo Classification Dataset CIFAR is an acronym that stand

ADITYA SHAH 1 Jan 05, 2022
Fast and simple implementation of RL algorithms, designed to run fully on GPU.

RSL RL Fast and simple implementation of RL algorithms, designed to run fully on GPU. This code is an evolution of rl-pytorch provided with NVIDIA's I

Robotic Systems Lab - Legged Robotics at ETH Zürich 68 Dec 29, 2022
HugsVision is a easy to use huggingface wrapper for state-of-the-art computer vision

HugsVision is an open-source and easy to use all-in-one huggingface wrapper for computer vision. The goal is to create a fast, flexible and user-frien

Labrak Yanis 166 Nov 27, 2022
The AugNet Python module contains functions for the fast computation of image similarity.

AugNet AugNet: End-to-End Unsupervised Visual Representation Learning with Image Augmentation arxiv link In our work, we propose AugNet, a new deep le

Ming 74 Dec 28, 2022
for a paper about leveraging discourse markers for training new models

TSLM-DISCOURSE-MARKERS Scope This repository contains: (1) Code to extract discourse markers from wikipedia (TSA). (1) Code to extract significant dis

International Business Machines 6 Nov 02, 2022
Replication Code for "Self-Supervised Bug Detection and Repair" NeurIPS 2021

Self-Supervised Bug Detection and Repair This is the reference code to replicate the research in Self-Supervised Bug Detection and Repair in NeurIPS 2

Microsoft 85 Dec 24, 2022
Edge-aware Guidance Fusion Network for RGB-Thermal Scene Parsing

EGFNet Edge-aware Guidance Fusion Network for RGB-Thermal Scene Parsing Dataset and Results Test maps: 百度网盘 提取码:zust Citation @ARTICLE{ author={Zhou,

ShaohuaDong 10 Dec 08, 2022
Simple image captioning model - CLIP prefix captioning.

CLIP prefix captioning. Inference Notebook: 🥳 New: 🥳 Our technical papar is finally out! Official implementation for the paper "ClipCap: CLIP Prefix

688 Jan 04, 2023
Tensorflow implementation of Human-Level Control through Deep Reinforcement Learning

Human-Level Control through Deep Reinforcement Learning Tensorflow implementation of Human-Level Control through Deep Reinforcement Learning. This imp

Devsisters Corp. 2.4k Dec 26, 2022
P-Tuning v2: Prompt Tuning Can Be Comparable to Finetuning Universally Across Scales and Tasks

P-tuning v2 P-Tuning v2: Prompt Tuning Can Be Comparable to Finetuning Universally Across Scales and Tasks An optimized prompt tuning strategy for sma

THUDM 540 Dec 30, 2022
Simple-System-Convert--C--F - Simple System Convert With Python

Simple-System-Convert--C--F REQUIREMENTS Python version : 3 HOW TO USE Run the c

Jonathan Santos 2 Feb 16, 2022
The Easy-to-use Dialogue Response Selection Toolkit for Researchers

Easy-to-use toolkit for retrieval-based Chatbot Recent Activity Our released RRS corpus can be found here. Our released BERT-FP post-training checkpoi

GMFTBY 32 Nov 13, 2022
Real Time Object Detection and Classification using Yolo Algorithm.

Real time Object detection & Classification using YOLO algorithm. Real Time Object Detection and Classification using Yolo Algorithm. What is Object D

Ketan Chawla 1 Apr 17, 2022
A fast MoE impl for PyTorch

An easy-to-use and efficient system to support the Mixture of Experts (MoE) model for PyTorch.

Rick Ho 873 Jan 09, 2023
Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods

Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods Introduction Graph Neural Networks (GNNs) have demonstrated

37 Dec 15, 2022