Revisiting Adversarial Training under Long-Tailed Distributions

• URL: http://arxiv.org/abs/2403.10073v1
• Date: Fri, 15 Mar 2024 07:29:41 GMT
• Title: Revisiting Adversarial Training under Long-Tailed Distributions
• Authors: Xinli Yue, Ningping Mou, Qian Wang, Lingchen Zhao,
• Abstract summary: Adversarial training has been recognized as one of the most effective methods to counter such attacks. We show that Balanced Softmax Loss alone can achieve performance comparable to the complete RoBal approach while significantly reducing training overheads. To address this, we explore data augmentation as a solution and unexpectedly discover that, unlike results obtained with balanced data, data augmentation not only effectively alleviates robust overfitting but also significantly improves robustness.
• Score: 8.187045490998269
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Deep neural networks are vulnerable to adversarial attacks, often leading to erroneous outputs. Adversarial training has been recognized as one of the most effective methods to counter such attacks. However, existing adversarial training techniques have predominantly been tested on balanced datasets, whereas real-world data often exhibit a long-tailed distribution, casting doubt on the efficacy of these methods in practical scenarios. In this paper, we delve into adversarial training under long-tailed distributions. Through an analysis of the previous work "RoBal", we discover that utilizing Balanced Softmax Loss alone can achieve performance comparable to the complete RoBal approach while significantly reducing training overheads. Additionally, we reveal that, similar to uniform distributions, adversarial training under long-tailed distributions also suffers from robust overfitting. To address this, we explore data augmentation as a solution and unexpectedly discover that, unlike results obtained with balanced data, data augmentation not only effectively alleviates robust overfitting but also significantly improves robustness. We further investigate the reasons behind the improvement of robustness through data augmentation and identify that it is attributable to the increased diversity of examples. Extensive experiments further corroborate that data augmentation alone can significantly improve robustness. Finally, building on these findings, we demonstrate that compared to RoBal, the combination of BSL and data augmentation leads to a +6.66% improvement in model robustness under AutoAttack on CIFAR-10-LT. Our code is available at https://github.com/NISPLab/AT-BSL .

Related papers

• Long-tailed Adversarial Training with Self-Distillation [15.184564265850382]
  We show that adversarial training struggles to achieve high performance on tail classes in long-tailed distributions. We propose a novel self-distillation technique to advance adversarial robustness on long-tailed distributions. Our experiments demonstrate state-of-the-art performance in both clean and robust accuracy for long-tailed adversarial robustness.
  arXiv  Detail & Related papers  (2025-03-09T05:39:36Z)
• TAET: Two-Stage Adversarial Equalization Training on Long-Tailed Distributions [3.9635480458924994]
  Adversarial robustness is a critical challenge in deploying deep neural networks for real-world applications. We propose a novel training framework, TAET, which integrates an initial stabilization phase followed by a stratified adversarial training phase. Our method surpasses existing advanced defenses, achieving significant improvements in both memory and computational efficiency.
  arXiv  Detail & Related papers  (2025-03-02T12:07:00Z)
• Improved Diffusion-based Generative Model with Better Adversarial Robustness [65.38540020916432]
  Diffusion Probabilistic Models (DPMs) have achieved significant success in generative tasks. During the denoising process, the input data distributions differ between the training and inference stages.
  arXiv  Detail & Related papers  (2025-02-24T12:29:16Z)
• Defending Against Sophisticated Poisoning Attacks with RL-based Aggregation in Federated Learning [12.352511156767338]
  Federated learning is highly susceptible to model poisoning attacks. In this paper, we propose AdaAggRL, an RL-based Adaptive aggregation method. Experiments on four real-world datasets demonstrate that the proposed defense model significantly outperforms widely adopted defense models for sophisticated attacks.
  arXiv  Detail & Related papers  (2024-06-20T11:33:14Z)
• Orthogonal Uncertainty Representation of Data Manifold for Robust Long-Tailed Learning [52.021899899683675]
  In scenarios with long-tailed distributions, the model's ability to identify tail classes is limited due to the under-representation of tail samples. We propose an Orthogonal Uncertainty Representation (OUR) of feature embedding and an end-to-end training strategy to improve the long-tail phenomenon of model robustness.
  arXiv  Detail & Related papers  (2023-10-16T05:50:34Z)
• Doubly Robust Instance-Reweighted Adversarial Training [107.40683655362285]
  We propose a novel doubly-robust instance reweighted adversarial framework. Our importance weights are obtained by optimizing the KL-divergence regularized loss function. Our proposed approach outperforms related state-of-the-art baseline methods in terms of average robust performance.
  arXiv  Detail & Related papers  (2023-08-01T06:16:18Z)
• Alleviating the Effect of Data Imbalance on Adversarial Training [26.36714114672729]
  We study adversarial training on datasets that obey the long-tailed distribution. We propose a new adversarial training framework -- Re-balancing Adversarial Training (REAT)
  arXiv  Detail & Related papers  (2023-07-14T07:01:48Z)
• A Data-Centric Approach for Improving Adversarial Training Through the Lens of Out-of-Distribution Detection [0.4893345190925178]
  We propose detecting and removing hard samples directly from the training procedure rather than applying complicated algorithms to mitigate their effects. Our results on SVHN and CIFAR-10 datasets show the effectiveness of this method in improving the adversarial training without adding too much computational cost.
  arXiv  Detail & Related papers  (2023-01-25T08:13:50Z)
• How Much Data Are Augmentations Worth? An Investigation into Scaling Laws, Invariance, and Implicit Regularization [76.58017437197859]
  We find that in out-of-distribution testing scenarios, augmentations which yield samples that are diverse, but inconsistent with the data distribution can be even more valuable than additional training data. We show that augmentations induce additionality during training, effectively flattening the loss landscape.
  arXiv  Detail & Related papers  (2022-10-12T17:42:01Z)
• Efficient Adversarial Training With Data Pruning [26.842714298874192]
  We show that data pruning leads to improvements in convergence and reliability of adversarial training. In some settings data pruning brings benefits from both worlds-it both improves adversarial accuracy and training time.
  arXiv  Detail & Related papers  (2022-07-01T23:54:46Z)
• Adversarial Robustness under Long-Tailed Distribution [93.50792075460336]
  Adversarial robustness has attracted extensive studies recently by revealing the vulnerability and intrinsic characteristics of deep networks. In this work we investigate the adversarial vulnerability as well as defense under long-tailed distributions. We propose a clean yet effective framework, RoBal, which consists of two dedicated modules, a scale-invariant and data re-balancing.
  arXiv  Detail & Related papers  (2021-04-06T17:53:08Z)
• Consistency Regularization for Adversarial Robustness [88.65786118562005]
  Adversarial training is one of the most successful methods to obtain the adversarial robustness of deep neural networks. However, a significant generalization gap in the robustness obtained from AT has been problematic. In this paper, we investigate data augmentation techniques to address the issue.
  arXiv  Detail & Related papers  (2021-03-08T09:21:41Z)
• Adversarial Self-Supervised Contrastive Learning [62.17538130778111]
  Existing adversarial learning approaches mostly use class labels to generate adversarial samples that lead to incorrect predictions. We propose a novel adversarial attack for unlabeled data, which makes the model confuse the instance-level identities of the perturbed data samples. We present a self-supervised contrastive learning framework to adversarially train a robust neural network without labeled data.
  arXiv  Detail & Related papers  (2020-06-13T08:24:33Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.