Enregistré dans:
Détails bibliographiques
Auteurs principaux: Wang, Longwei, Nayyem, Mohammad Navid, Rakin, Abdullah Al, Santosh, KC, Zhang, Chaowei, Zhou, Yang
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:https://arxiv.org/abs/2601.00968
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911350975889408
author Wang, Longwei
Nayyem, Mohammad Navid
Rakin, Abdullah Al
Santosh, KC
Zhang, Chaowei
Zhou, Yang
author_facet Wang, Longwei
Nayyem, Mohammad Navid
Rakin, Abdullah Al
Santosh, KC
Zhang, Chaowei
Zhou, Yang
contents The growing reliance on deep learning models in safety-critical domains such as healthcare and autonomous navigation underscores the need for defenses that are both robust to adversarial perturbations and transparent in their decision-making. In this paper, we identify a connection between interpretability and robustness that can be directly leveraged during training. Specifically, we observe that spurious, unstable, or semantically irrelevant features identified through Local Interpretable Model-Agnostic Explanations (LIME) contribute disproportionately to adversarial vulnerability. Building on this insight, we introduce an attribution-guided refinement framework that transforms LIME from a passive diagnostic into an active training signal. Our method systematically suppresses spurious features using feature masking, sensitivity-aware regularization, and adversarial augmentation in a closed-loop refinement pipeline. This approach does not require additional datasets or model architectures and integrates seamlessly into standard adversarial training. Theoretically, we derive an attribution-aware lower bound on adversarial distortion that formalizes the link between explanation alignment and robustness. Empirical evaluations on CIFAR-10, CIFAR-10-C, and CIFAR-100 demonstrate substantial improvements in adversarial robustness and out-of-distribution generalization.
format Preprint
id arxiv_https___arxiv_org_abs_2601_00968
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Explainability-Guided Defense: Attribution-Aware Model Refinement Against Adversarial Data Attacks
Wang, Longwei
Nayyem, Mohammad Navid
Rakin, Abdullah Al
Santosh, KC
Zhang, Chaowei
Zhou, Yang
Machine Learning
The growing reliance on deep learning models in safety-critical domains such as healthcare and autonomous navigation underscores the need for defenses that are both robust to adversarial perturbations and transparent in their decision-making. In this paper, we identify a connection between interpretability and robustness that can be directly leveraged during training. Specifically, we observe that spurious, unstable, or semantically irrelevant features identified through Local Interpretable Model-Agnostic Explanations (LIME) contribute disproportionately to adversarial vulnerability. Building on this insight, we introduce an attribution-guided refinement framework that transforms LIME from a passive diagnostic into an active training signal. Our method systematically suppresses spurious features using feature masking, sensitivity-aware regularization, and adversarial augmentation in a closed-loop refinement pipeline. This approach does not require additional datasets or model architectures and integrates seamlessly into standard adversarial training. Theoretically, we derive an attribution-aware lower bound on adversarial distortion that formalizes the link between explanation alignment and robustness. Empirical evaluations on CIFAR-10, CIFAR-10-C, and CIFAR-100 demonstrate substantial improvements in adversarial robustness and out-of-distribution generalization.
title Explainability-Guided Defense: Attribution-Aware Model Refinement Against Adversarial Data Attacks
topic Machine Learning
url https://arxiv.org/abs/2601.00968