Weakly Supervised Contrastive Adversarial Training for Learning Robust Features from Semi-supervised Data
Not on the map. Why?
Abstract (arXiv)
Existing adversarial training (AT) methods often suffer from incomplete perturbation, meaning that not all non-robust features are perturbed when generating adversarial examples (AEs). This results in residual correlations between non-robust features and labels, leading to suboptimal learning of robust features. However, achieving complete perturbation, i.e., perturbing as many non-robust features as possible, is challenging due to the difficulty in distinguishing robust and non-robust features and the sparsity of labeled data. To address these challenges, we propose a novel approach called Weakly Supervised Contrastive Adversarial Training (WSCAT). WSCAT ensures complete perturbation for improved learning of robust features by disrupting correlations between non-robust features and labels through complete AE generation over partially labeled data, grounded in information theory. Extensive theoretical analysis and comprehensive experiments on widely adopted benchmarks validate the superiority of WSCAT. Our code is available at https://github.com/zhang-lilin/WSCAT.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 30 of its 47 references that are in the map. First posted on arXiv 14 Mar 2025.
Builds on
30 of 47 references.
References · 30
- 2014Explaining and Harnessing Adversarial Examplescited by 7,287
- 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
- 2013Intriguing properties of neural networkscited by 5,114
- 2016Wide Residual Networkscited by 3,057
- 2016Towards Evaluating the Robustness of Neural Networkscited by 2,976
- 2020Supervised Contrastive Learningcited by 2,803
- 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,510
- 2020Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attackscited by 1,442
- 2019Adversarial Examples Are Not Bugs, They Are Featurescited by 950
- 2017Evasion Attacks against Machine Learning at Test Timecited by 686
- 2019On Variational Bounds of Mutual Informationcited by 550
- 2020Overfitting in adversarially robust deep learningcited by 504
18 more
- 2019Unlabeled Data Improves Adversarial Robustnesscited by 415
- 2017A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasetscited by 374
- 2021Recent Advances in Adversarial Training for Adversarial Robustnesscited by 276
- 2023Better Diffusion Models Further Improve Adversarial Trainingcited by 207
- 2020Theoretical Analysis of Self-Training with Deep Networks on Unlabeled Datacited by 179
- 2019Are Labels Required for Improving Adversarial Robustness?cited by 162
- 2019Defense Against Adversarial Attacks Using Feature Scattering-based Adversarial Trainingcited by 95
- 2021CausalAdv: Adversarial Robustness through the Lens of Causalitycited by 37
- 2024Revisiting Adversarial Training at Scalecited by 33
- 2019Robust Local Features for Improving the Generalization of Adversarial Trainingcited by 30
- 2022Distilling Robust and Non-Robust Features in Adversarial Examples by Information Bottleneckcited by 30
- 2023Demystifying Causal Features on Adversarial Examples and Causal Inoculation for Robust Network by Adversarial Instrumental Variable Regressioncited by 18
- 2022InfoAT: Improving Adversarial Training Using the Information Bottleneck Principlecited by 7
- 2023Provable Unrestricted Adversarial Training without Compromise with Generalizabilitycited by 6
- 2023Focus on Hiders: Exploring Hidden Threats for Enhancing Adversarial Trainingcited by 5
- 2023Exploring Robust Features for Improving Adversarial Robustnesscited by 4
- 2023Robust Proxy: Improving Adversarial Robustness by Robust Proxy Learningcited by 3
- 2023Where and How to Attack? A Causality-Inspired Recipe for Generating Counterfactual Adversarial Examplescited by 3
Built on by
Citing papers per year: 2026: 1.
1 paper.