EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Mixed Adversarial Awareness Technique for Improving Neural Network Defense

Moses Apambila Agebure, Sampson Konja, Stephen Akobre, Mohammed Daabo

Journal of Information Systems and Informatics · 26.06.2026 · 8 (3) · 3623-3643

DOI: 10.63158/journalisi.v8i3.1552 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Information Systems and Informatics, online erschienen am 26.06.2026. Verfasst von Moses Apambila Agebure und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Neural Network (NN) models, particularly Convolutional Neural Networks (CNNs), have achieved remarkable performance in computer vision tasks but remain highly vulnerable to adversarial attacks. Existing defense techniques mainly focus on detecting adversarial examples and often show limited effectiveness when adversarial perturbations coexist with significant noisy inputs. To address this limitation, this study proposes a Mixed Adversarial Awareness Technique (MAAT) based on kernel density estimation and a Bayesian uncertainty estimator. Kernel density estimation is used to model data manifolds in the input subspace, while the Bayesian uncertainty estimator, inspired by the Dirichlet process, quantifies predictive uncertainty in the input space. The proposed technique was evaluated on three benchmark datasets, CIFAR-10, CIFAR-100, and SVHN, using four adversarial attack schemes, namely FGSM, BIM, JSMA, and C&W, as well as Gaussian noise injection. The LeNet ConvNet model was employed as the test classifier. Experimental results show that MAAT effectively flags adversarial and noisy instances, improving detection performance with AUC values ranging from 0.84 to 0.96, compared with 0.61 to 0.94 achieved by selected state-of-the-art techniques. These findings demonstrate that combining density-based manifold modeling with uncertainty estimation provides a robust defense against mixed adversarial and noisy inputs.

Zitieren

APA

Agebure, M. A., Konja, S., Akobre, S., & Daabo, M. (2026). A Mixed Adversarial Awareness Technique for Improving Neural Network Defense. Journal of Information Systems and Informatics, 8(3), 3623-3643. https://doi.org/10.63158/journalisi.v8i3.1552

Vancouver

Agebure MA, Konja S, Akobre S, Daabo M. A Mixed Adversarial Awareness Technique for Improving Neural Network Defense. Journal of Information Systems and Informatics. 2026;8(3):3623-3643. doi:10.63158/journalisi.v8i3.1552

BibTeX
@article{Agebure2026,
  author = {Agebure, Moses Apambila and Konja, Sampson and Akobre, Stephen and Daabo, Mohammed},
  title = {A Mixed Adversarial Awareness Technique for Improving Neural Network Defense},
  journal = {Journal of Information Systems and Informatics},
  year = {2026},
  volume = {8},
  number = {3},
  pages = {3623--3643},
  doi = {10.63158/journalisi.v8i3.1552}
}
Kostenlos starten →