Bounded Adversarial Attack on Deep Content Features
Abstract
We propose a novel adversarial attack targeting content features in some deep layer, that is, individual neurons in the layer. A naive method that enforces a fixed value/percentage bound for neuron activation values can hardly work and generates very noisy samples. The reason is that the level of perceptual variation entailed by a fixed value bound is non-uniform across neurons and even for the same neuron. We hence propose a novel distribution quantile bound for activation values and a polynomial barrier loss function. Given a benign input, a fixed quantile bound is translated to many value bounds, one for each neuron, based on the distributions of the neuron's activations and the current activation value on the given input. These individualized bounds enable fine-grained regulation, allowing content feature mutations with bounded perceptional variations. Our evaluation on ImageNet and five different model architectures demonstrates that our attack is effective. Compared to seven other latest adversarial attacks in both the pixel space and the feature space, our attack can achieve the state-of-the-art trade-off between attack success rate and imperceptibility. <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> Code and Samples are available on Github [37].
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.
1 top-tier paper has cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Feature Squeezing: Detecting Adversarial Examples in Deep Neural Networks cites Towards Evaluating the Robustness of Neural Networks
- Sparse and Imperceivable Adversarial Attacks cites Towards Evaluating the Robustness of Neural Networks
- On Adaptive Attacks to Adversarial Example Defenses cites Towards Evaluating the Robustness of Neural Networks
- Unrestricted Adversarial Examples via Semantic Manipulation cites Towards Evaluating the Robustness of Neural Networks
- Unrestricted Adversarial Examples via Semantic Manipulation cites Feature Squeezing: Detecting Adversarial Examples in Deep Neural Networks
Cited by top-tier papers1
Builds on6
- Towards Evaluating the Robustness of Neural NetworksNicholas Carlini, David A. WagnerS&P 2017 · 9,786 citations
- Feature Squeezing: Detecting Adversarial Examples in Deep Neural NetworksWeilin Xu, David Evans, Yanjun QiNDSS 2018 · 1,633 citations
- On Adaptive Attacks to Adversarial Example DefensesFlorian Tramèr, Nicholas Carlini, Wieland Brendel, Aleksander MadryNeurIPS 2020 · 1,026 citations
- Sparse and Imperceivable Adversarial AttacksFrancesco Croce, Matthias HeinICCV 2019 · 228 citations
- Unrestricted Adversarial Examples via Semantic ManipulationAnand Bhattad, Min Jin Chong, Kaizhao Liang, Bo Li et al.ICLR 2020 · 177 citations
Related papers
- Towards Feature Space Adversarial Attack by Style PerturbationQiuling Xu, Guanhong Tao, Siyuan Cheng, Xiangyu ZhangAAAI 2021 · 33 citations
- Content-based Unrestricted Adversarial AttackZhaoyu Chen, Bo Li, Shuang Wu, Kaixun Jiang et al.NeurIPS 2023 · 132 citations
- StyLess: Boosting the Transferability of Adversarial ExamplesKaisheng Liang, Bin XiaoCVPR 2023
- ColorFool: Semantic Adversarial ColorizationAli Shahin Shamsabadi, Ricardo Sánchez-Matilla, Andrea CavallaroCVPR 2020
- AdvAD: Exploring Non-Parametric Diffusion for Imperceptible Adversarial AttacksJin Li, Ziqiang He, Anwei Luo, Jian-Fang Hu et al.NeurIPS 2024 · 17 citations
