Lune

CVPR2020Top-tier venue

Polishing Decision-Based Adversarial Noise With a Customized Sampling

Yucheng Shi, Yahong Han, Qi Tian

Abstract

As an effective black-box adversarial attack, decisionbased methods polish adversarial noise by querying the target model. Among them, boundary attack is widely applied due to its powerful noise compression capability, especially when combined with transfer-based methods. Boundary attack splits the noise compression into several independent sampling processes, repeating each query with a constant sampling setting. In this paper, we demonstrate the advantage of using current noise and historical queries to customize the variance and mean of sampling in boundary attack to polish adversarial noise. We further reveal the relationship between the initial noise and the compressed noise in boundary attack. We propose Customized Adversarial Boundary (CAB) attack that uses the current noise to model the sensitivity of each pixel and polish adversarial noise of each image with a customized sampling setting. On the one hand, CAB uses current noise as a prior belief to customize the multivariate normal distribution. On the other hand, CAB keeps the new samplings away from historical failed queries to avoid similar mistakes. Experimental results measured on several image classification datasets emphasizes the validity of our method.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

3 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • The LogBarrier Adversarial Attack: Making Effective Use of Decision Boundary Information cites Towards Evaluating the Robustness of Neural Networks
  • Finding Optimal Tangent Points for Reducing Distortions of Hard-label Attacks cites Towards Evaluating the Robustness of Neural Networks
  • Finding Optimal Tangent Points for Reducing Distortions of Hard-label Attacks cites Guessing Smart: Biased Sampling for Efficient Black-Box Adversarial Attacks
  • Aha! Adaptive History-driven Attack for Decision-based Black-box Models cites Towards Evaluating the Robustness of Neural Networks
  • Aha! Adaptive History-driven Attack for Decision-based Black-box Models cites Guessing Smart: Biased Sampling for Efficient Black-Box Adversarial Attacks
  • Aha! Adaptive History-driven Attack for Decision-based Black-box Models cites HopSkipJumpAttack: A Query-Efficient Decision-Based Attack
  • Guessing Smart: Biased Sampling for Efficient Black-Box Adversarial Attacks cites Towards Evaluating the Robustness of Neural Networks
  • Decision-based Black-box Attack Against Vision Transformers via Patch-wise Adversarial Removal cites Guessing Smart: Biased Sampling for Efficient Black-Box Adversarial Attacks
  • Decision-based Black-box Attack Against Vision Transformers via Patch-wise Adversarial Removal cites HopSkipJumpAttack: A Query-Efficient Decision-Based Attack
  • HopSkipJumpAttack: A Query-Efficient Decision-Based Attack cites Towards Evaluating the Robustness of Neural Networks
  • HopSkipJumpAttack: A Query-Efficient Decision-Based Attack cites Guessing Smart: Biased Sampling for Efficient Black-Box Adversarial Attacks
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers3

  1. Decision-based Black-box Attack Against Vision Transformers via Patch-wise Adversarial RemovalYucheng Shi, Yahong Han, Yu-an Tan, Xiaohui KuangNeurIPS 2022 · 43 citations
  2. Aha! Adaptive History-driven Attack for Decision-based Black-box ModelsJie Li, Rongrong Ji, Peixian Chen, Baochang Zhang et al.ICCV 2021 · 25 citations
  3. Finding Optimal Tangent Points for Reducing Distortions of Hard-label AttacksChen Ma, Xiangyu Guo, Li Chen, Jun-Hai Yong et al.NeurIPS 2021 · 24 citations

Builds on4

  1. Towards Evaluating the Robustness of Neural NetworksNicholas Carlini, David A. WagnerS&P 2017 · 9,786 citations
  2. HopSkipJumpAttack: A Query-Efficient Decision-Based AttackJianbo Chen, Michael I. Jordan, Martin J. WainwrightS&P 2020 · 797 citations
  3. Guessing Smart: Biased Sampling for Efficient Black-Box Adversarial AttacksThomas Brunner, Frederik Diehl, Michael Truong-Le, Alois C. KnollICCV 2019 · 127 citations
  4. The LogBarrier Adversarial Attack: Making Effective Use of Decision Boundary InformationChris Finlay, Aram-Alexandre Pooladian, Adam M. ObermanICCV 2019 · 30 citations

Related papers

  1. ADBA: Approximation Decision Boundary Approach for Black-Box Adversarial AttacksFeiyang Wang, Xingquan Zuo, Hai Huang, Gang ChenAAAI 2025 · 14 citations
  2. CGBA: Curvature-aware Geometric Black-box AttackMd Farhamdur Reza, Ali Rahmati, Tianfu Wu, Huaiyu DaiICCV 2023 · 33 citations
  3. Delving into Deep Image Prior for Adversarial Defense: A Novel Reconstruction-based Defense FrameworkLi Ding, Yongwei Wang, Xin Ding, Kaiwen Yuan et al.ACM MM 2021 · 7 citations
  4. A Geometry-Inspired Decision-Based AttackYujia Liu, Seyed-Mohsen Moosavi-Dezfooli, Pascal FrossardICCV 2019 · 55 citations
  5. Transferability of White-box Perturbations: Query-Efficient Adversarial Attacks against Commercial DNN ServicesMeng Shen, Changyue Li, Qi Li, Hao Lu et al.USENIX Security 2024 · 8 citations
Dusk over the sea between two cliffs drawn in fine vertical lines