EUVIMEDEUVIMED entdecken
Crossref

PadNet: Defending Neural Networks Against Adversarial Examples

Armon Barton, Matthew Wright, Shaikh Akib Shahriyar, Edgar Jatho, Mohammad Saidur Rahman, Kantha Girish Gangadhara, Jiang Ming

ACM Transactions on Privacy and Security · 31.05.2026 · 29 (2) · 1-26

DOI: 10.1145/3799889 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ACM Transactions on Privacy and Security, online erschienen am 24.03.2026. Verfasst von Armon Barton und 6 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben. Gefördert von: National Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Machine learning (ML) suffers from a persistent and critical flaw: adversarial examples. Many new forms of adversarial example attacks have been invented and many narrow defenses have been proposed. Unfortunately, no defensive approach can withstand current attacks. We hypothesize that ML model robustness can be improved with approaches that delineate the data-point-sparse latent space between data-dense regions of a model’s classification space as a barrier class. We introduce one such defense, PadNet, that builds a barrier class using a combination of training samples that mix multiple classes together. It leverages this barrier class to separate decision boundaries between benign classes with regions of padding. PadNet then implements a gradient regularization strategy that penalizes gradients in the direction of the barrier class, causing the decision boundary to draw tighter around training samples increasing boundary thickness between classes. We evaluate PadNet against a sampling of the most effective state-of-the-art attacks, demonstrating that it offers significant robustness and reliability compared to current defenses. We also test it against adaptive attacks and find that PadNet remains robust against them.

Förderung

  • National Science Foundation

Zitieren

APA

Barton, A., Wright, M., Shahriyar, S. A., Jatho, E., Rahman, M. S., Gangadhara, K. G., & Ming, J. (2026). PadNet: Defending Neural Networks Against Adversarial Examples. ACM Transactions on Privacy and Security, 29(2), 1-26. https://doi.org/10.1145/3799889

Vancouver

Barton A, Wright M, Shahriyar SA, Jatho E, Rahman MS, Gangadhara KG, et al. PadNet: Defending Neural Networks Against Adversarial Examples. ACM Transactions on Privacy and Security. 2026;29(2):1-26. doi:10.1145/3799889

BibTeX
@article{Barton2026,
  author = {Barton, Armon and Wright, Matthew and Shahriyar, Shaikh Akib and Jatho, Edgar and Rahman, Mohammad Saidur and Gangadhara, Kantha Girish and Ming, Jiang},
  title = {PadNet: Defending Neural Networks Against Adversarial Examples},
  journal = {ACM Transactions on Privacy and Security},
  year = {2026},
  volume = {29},
  number = {2},
  pages = {1--26},
  doi = {10.1145/3799889}
}
Kostenlos starten →