Timezone: »
We propose functional adversarial attacks, a novel class of threat models for crafting adversarial examples to fool machine learning models. Unlike a standard lp-ball threat model, a functional adversarial threat model allows only a single function to be used to perturb input features to produce an adversarial example. For example, a functional adversarial attack applied on colors of an image can change all red pixels simultaneously to light red. Such global uniform changes in images can be less perceptible than perturbing pixels of the image individually. For simplicity, we refer to functional adversarial attacks on image colors as ReColorAdv, which is the main focus of our experiments. We show that functional threat models can be combined with existing additive (lp) threat models to generate stronger threat models that allow both small, individual perturbations and large, uniform changes to an input. Moreover, we prove that such combinations encompass perturbations that would not be allowed in either constituent threat model. In practice, ReColorAdv can significantly reduce the accuracy of a ResNet-32 trained on CIFAR-10. Furthermore, to the best of our knowledge, combining ReColorAdv with other attacks leads to the strongest existing attack even after adversarial training.
Author Information
Cassidy Laidlaw (University of Maryland, College Park)
Soheil Feizi (University of Maryland)
More from the Same Authors
-
2021 Spotlight: Uncertain Decisions Facilitate Better Preference Learning »
Cassidy Laidlaw · Stuart Russell -
2022 Poster: Hard ImageNet: Segmentations for Objects with Strong Spurious Cues »
Mazda Moayeri · Sahil Singla · Soheil Feizi -
2022 Poster: Explicit Tradeoffs between Adversarial and Natural Distributional Robustness »
Mazda Moayeri · Kiarash Banihashem · Soheil Feizi -
2022 Poster: Lethal Dose Conjecture on Data Poisoning »
Wenxiao Wang · Alexander Levine · Soheil Feizi -
2022 Poster: Toward Efficient Robust Training against Union of $\ell_p$ Threat Models »
Gaurang Sriramanan · Maharshi Gor · Soheil Feizi -
2022 Poster: Improved techniques for deterministic l2 robustness »
Sahil Singla · Soheil Feizi -
2021 Poster: Uncertain Decisions Facilitate Better Preference Learning »
Cassidy Laidlaw · Stuart Russell -
2021 Poster: Improving Deep Learning Interpretability by Saliency Guided Training »
Aya Abdelsalam Ismail · Hector Corrada Bravo · Soheil Feizi -
2020 : Opening Remarks »
Reinhard Heckel · Paul Hand · Soheil Feizi · Lenka Zdeborová · Richard Baraniuk -
2020 Workshop: Workshop on Deep Learning and Inverse Problems »
Reinhard Heckel · Paul Hand · Richard Baraniuk · Lenka Zdeborová · Soheil Feizi -
2020 Poster: Certifying Confidence via Randomized Smoothing »
Aounon Kumar · Alexander Levine · Soheil Feizi · Tom Goldstein -
2020 Poster: Robust Optimal Transport with Applications in Generative Modeling and Domain Adaptation »
Yogesh Balaji · Rama Chellappa · Soheil Feizi -
2020 Poster: Dual Manifold Adversarial Robustness: Defense against Lp and non-Lp Adversarial Attacks »
Wei-An Lin · Chun Pong Lau · Alexander Levine · Rama Chellappa · Soheil Feizi -
2020 Poster: Benchmarking Deep Learning Interpretability in Time Series Predictions »
Aya Abdelsalam Ismail · Mohamed Gunady · Hector Corrada Bravo · Soheil Feizi -
2020 Poster: (De)Randomized Smoothing for Certifiable Defense against Patch Attacks »
Alexander Levine · Soheil Feizi -
2019 : Soheil Feizi, "Certifiable Defenses against Adversarial Attacks" »
Soheil Feizi -
2019 Poster: Quantum Wasserstein Generative Adversarial Networks »
Shouvanik Chakrabarti · Huang Yiming · Tongyang Li · Soheil Feizi · Xiaodi Wu -
2019 Poster: Input-Cell Attention Reduces Vanishing Saliency of Recurrent Neural Networks »
Aya Abdelsalam Ismail · Mohamed Gunady · Luiz Pessoa · Hector Corrada Bravo · Soheil Feizi -
2018 Poster: Porcupine Neural Networks: Approximating Neural Network Landscapes »
Soheil Feizi · Hamid Javadi · Jesse Zhang · David Tse -
2017 Poster: Tensor Biclustering »
Soheil Feizi · Hamid Javadi · David Tse -
2014 Poster: Biclustering Using Message Passing »
Luke O'Connor · Soheil Feizi