Attack classification models with transferability, black-box attack; unrestricted adversarial attacks on imagenet

Overview

Attack_classification_models_with_transferability

Attack classification models with transferability, black-box attack; unrestricted adversarial attacks on imagenet, CVPR2021 安全AI挑战者计划第六期:ImageNet无限制对抗攻击, 决赛第四名(team name: Advers)

详细方案介绍

1. Prerequisites

1. python >= 3.6
2. pytorch >= 1.2.0
3. torchvision >= 0.4.0 
4. numpy >= 1.19.1
5. pillow >= 7.2.0
6. scipy >= 1.5.2

2. Code Overview

  • ./codes/

    • main.py: 攻击原始图像,生成并保存攻击后的图像
    • data.py: 加载原始图像,保存图像,图像标准化处理
    • model.py: 模型集成,利用集成模型计算logits
    • utils.py: Input Diversity, 高斯平滑处理等
  • ./input_dir/

    • ./images/: 原始图像所在路径
    • ./dev.csv:图像的标记文件(images name, true label)
  • demo

python main.py --source_model 'resnet50'

3. 思路

  • 本文分享我们团队(Advers)的解决方案,欢迎大家交流讨论,一起进步。
  • 本方案最终得分:9081.6, 线上后台模型攻击成功率:95.48%,决赛排名:TOP 4
  • 本方案初赛排名:TOP 4,复赛排名:TOP 10

3.1 赛题分析

  1. 无限制对抗攻击可以用不同的方法来实现,包括范数扰动攻击、GAN、粘贴Patch等。但是由于 fid、lpips 两个指标的限制,必须保证生成的图像质量好(不改变语义、噪声尽量小),否则得分会很低。经过尝试,我们最终确定利用范数扰动来进行迁移攻击,这样可以较好地平衡攻击成功率和图像质量。

  2. 由于无法获取后台模型的任何参数和输出,甚至不知道后台分类模型输入图像大小,这增加了攻击难度。原图大小是 500 * 500,而 ImageNet 分类模型输入是 224 * 224 或 299 * 299,对生成的对抗样本图像 resize会导致对抗样本的攻击性降低。

  3. 由于比赛最终排名为人工打分,所以没有用损失函数去拟合 fid、lpips 两个指标。

  4. 对抗样本的攻击性和图像质量可以说是两个相互矛盾的指标,一个指标的提升往往会导致另一个指标的下降,如何在对抗性和图像质量两个方面找到一个平衡点是十分重要的。在机器打分阶段,采用较小的噪声,把噪声加在图像敏感区域,在尽量不降低攻击性的前提下提升对抗样本的图像质量是得分的关键

3.2 解题思路

3.2.1 输入模型的图像大小

本次比赛的图像被 resize 到了 500 * 500 大小,而标准的 ImageNet 预训练模型输入大小一般是 224 * 224 或 299 * 299。我们尝试将不同大小的图片(500,299,224)输入到模型中进行攻击,发现 224 大小的效果最好,计算复杂度也最低。

3.2.2 L2 or Linf

采用 L2 范数攻击生成的对抗样本的攻击性要强一些,但可能会出现比较大的噪声斑块,导致人眼看起来比较奇怪,采用 Linf 范数生成的对抗样本,人眼视觉上要稍好一些。在机器打分阶段,采用 L2 范数扰动攻击,在人工评判阶段,采用 Linf 范数扰动来生成对抗样本。

3.2.3 提升对抗样本迁移性方法

1. MI-FGSM1:在机器打分阶段采用 MI-FGSM 算法生成噪声,但是 MI-FGSM 算法生成的噪声人眼看起来会明显,由于决赛阶段是人工打分,最终舍弃了该方法。

2. Translation-Invariant(TI)2:用核函数对计算得到的噪声梯度进行平滑处理,提升了噪声的泛化性。

3. Input Diversity(DI)3 :通过增加输入图像的多样性来提高对抗样本的迁移性,其提分效果明显。Input Diversity 本质是通过变换输入图像的多样性让噪声不完全依赖相应的像素点,减少了噪声过拟合效应,提高了泛化性和迁移性。

3.2.4 改进后的DI攻击

Input Diversity 会对图像进行随机变换,导致生成的噪声梯度带有一定的随机性。虽然这种随机性可以使对抗样本的泛化性更强,但是也会引入一定比例的噪声,这种噪声也会抑制对抗样本的泛化性,因此如何消除 DI 随机性带来的噪声影响,同时保证攻击具有较强的泛化性是提升迁移性的有效手段。

image

3.2.5 Tricks

  • 在初赛和复赛阶段,采用 L2 和 Linf 范数扰动攻击,其中 L2 范数扰动攻击得分更高一些。由于复赛阶段线上模型比较鲁棒,所以适当增加扰动范围是提升攻击成功率的关键。
  • 考虑到决赛阶段是人工打分,需要考虑攻击性和图像质量,我们最终采用 Linf 范数扰动进行攻击,扰动大小设为 32/255,迭代次数设为 40,迭代步长设为 1/255。
  • 攻击之前,对图像进行高斯平滑处理,可以提升攻击效果,但是也会让图像变模糊。
  • Ensemble models: resnet50、densenet161、inceptionv4等。

4. 攻击结果

image

多次实验表明,采用改进的 DI+TI 攻击方法得到的噪声相对于 MI-FGSM 方法更小,泛化性和迁移性更强,同时人眼视觉效果也比较好。

5. 参考文献

  1. Dong Y, Liao F, Pang T, et al. Boosting adversarial attacks with momentum. CVPR 2018.
  2. Dong Y, Pang T, Su H, et al. Evading defenses to transferable adversarial examples by translation-invariant attacks. CVPR 2019.
  3. Xie C, Zhang Z, Zhou Y, et al. Improving transferability of adversarial examples with input diversity. CVPR 2019.
  4. Wierstra D, Schaul T, Glasmachers T, et al. Natural evolution strategies. The Journal of Machine Learning Research, 2014.

6.致谢

  • 感谢团队每一位小伙伴的辛勤付出,感谢指导老师的大力支持。
  • 感谢阿里安全主办了这次比赛,给了大家交流学习的机会,使得我们结识了很多优秀的小伙伴!

如有问题,欢迎交流:[email protected]

DWIPrep is a robust and easy-to-use pipeline for preprocessing of diverse dMRI data.

DWIPrep: A Robust Preprocessing Pipeline for dMRI Data DWIPrep is a robust and easy-to-use pipeline for preprocessing of diverse dMRI data. The transp

Gal Ben-Zvi 1 Jan 09, 2023
Transformer Huffman coding - Complete Huffman coding through transformer

Transformer_Huffman_coding Complete Huffman coding through transformer 2022/2/19

3 May 19, 2022
Proto-RL: Reinforcement Learning with Prototypical Representations

Proto-RL: Reinforcement Learning with Prototypical Representations This is a PyTorch implementation of Proto-RL from Reinforcement Learning with Proto

Denis Yarats 74 Dec 06, 2022
Inferred Model-based Fuzzer

IMF: Inferred Model-based Fuzzer IMF is a kernel API fuzzer that leverages an automated API model inferrence techinque proposed in our paper at CCS. I

SoftSec Lab 104 Sep 28, 2022
Implementation of paper "Decision-based Black-box Attack Against Vision Transformers via Patch-wise Adversarial Removal"

Patch-wise Adversarial Removal Implementation of paper "Decision-based Black-box Attack Against Vision Transformers via Patch-wise Adversarial Removal

4 Oct 12, 2022
Adversarial examples to the new ConvNeXt architecture

Adversarial examples to the new ConvNeXt architecture To get adversarial examples to the ConvNeXt architecture, run the Colab: https://github.com/stan

Stanislav Fort 19 Sep 18, 2022
cisip-FIRe - Fast Image Retrieval

Fast Image Retrieval (FIRe) is an open source image retrieval project release by Center of Image and Signal Processing Lab (CISiP Lab), Universiti Malaya. This project implements most of the major bi

CISiP Lab 39 Nov 25, 2022
This project deploys a yolo fastest model in the form of tflite on raspberry 3b+. The model is from another repository of mine called -Trash-Classification-Car

Deploy-yolo-fastest-tflite-on-raspberry 觉得有用的话可以顺手点个star嗷 这个项目将垃圾分类小车中的tflite模型移植到了树莓派3b+上面。 该项目主要是为了记录在树莓派部署yolo fastest tflite的流程 (之后有时间会尝试用C++部署来提升

7 Aug 16, 2022
ICCV2021: Code for 'Spatial Uncertainty-Aware Semi-Supervised Crowd Counting'

ICCV2021: Code for 'Spatial Uncertainty-Aware Semi-Supervised Crowd Counting'

Yanda Meng 14 May 13, 2022
Open source implementation of "A Self-Supervised Descriptor for Image Copy Detection" (SSCD).

A Self-Supervised Descriptor for Image Copy Detection (SSCD) This is the open-source codebase for "A Self-Supervised Descriptor for Image Copy Detecti

Meta Research 68 Jan 04, 2023
SCAN: Learning to Classify Images without Labels, incl. SimCLR. [ECCV 2020]

Learning to Classify Images without Labels This repo contains the Pytorch implementation of our paper: SCAN: Learning to Classify Images without Label

Wouter Van Gansbeke 1.1k Dec 30, 2022
A PyTorch-based Semi-Supervised Learning (SSL) Codebase for Pixel-wise (Pixel) Vision Tasks

PixelSSL is a PyTorch-based semi-supervised learning (SSL) codebase for pixel-wise (Pixel) vision tasks. The purpose of this project is to promote the

Zhanghan Ke 255 Dec 11, 2022
BossNAS: Exploring Hybrid CNN-transformers with Block-wisely Self-supervised Neural Architecture Search

BossNAS This repository contains PyTorch evaluation code, retraining code and pretrained models of our paper: BossNAS: Exploring Hybrid CNN-transforme

Changlin Li 127 Dec 26, 2022
Probabilistic Gradient Boosting Machines

PGBM Probabilistic Gradient Boosting Machines (PGBM) is a probabilistic gradient boosting framework in Python based on PyTorch/Numba, developed by Air

Olivier Sprangers 112 Dec 28, 2022
Code for the CVPR 2021 paper: Understanding Failures of Deep Networks via Robust Feature Extraction

Welcome to Barlow Barlow is a tool for identifying the failure modes for a given neural network. To achieve this, Barlow first creates a group of imag

Sahil Singla 33 Dec 05, 2022
TensorFlow implementation of "TokenLearner: What Can 8 Learned Tokens Do for Images and Videos?"

TokenLearner: What Can 8 Learned Tokens Do for Images and Videos? Source: Improving Vision Transformer Efficiency and Accuracy by Learning to Tokenize

Aritra Roy Gosthipaty 23 Dec 24, 2022
Modifications of the official PyTorch implementation of StyleGAN3. Let's easily generate images and videos with StyleGAN2/2-ADA/3!

Alias-Free Generative Adversarial Networks (StyleGAN3) Official PyTorch implementation of the NeurIPS 2021 paper Alias-Free Generative Adversarial Net

Diego Porres 185 Dec 24, 2022
A Simple LSTM-Based Solution for "Heartbeat Signal Classification and Prediction" in Tianchi

LSTM-Time-Series-Prediction A Simple LSTM-Based Solution for "Heartbeat Signal Classification and Prediction" in Tianchi Contest. The Link of the Cont

KevinCHEN 1 Jun 13, 2022
PyQt6 configuration in yaml format providing the most simple script.

PyamlQt(ぴゃむるきゅーと) PyQt6 configuration in yaml format providing the most simple script. Requirements yaml PyQt6, ( PyQt5 ) Installation pip install Pya

Ar-Ray 7 Aug 15, 2022
A diff tool for language models

LMdiff Qualitative comparison of large language models. Demo & Paper: http://lmdiff.net LMdiff is a MIT-IBM Watson AI Lab collaboration between: Hendr

Hendrik Strobelt 27 Dec 29, 2022