EUVIMEDEUVIMED entdecken
Crossref

Adversarial Machine Learning in Cybersecurity: A Comprehensive Framework for Threat Detection and Mitigation Using Deep Neural Networks

Nilesh Parihar

Dandao Xuebao/Journal of Ballistics · 21.05.2026 · 38 (1) · 391-399

DOI: 10.52783/dxjb.v38.288 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Dandao Xuebao/Journal of Ballistics, online erschienen am 21.05.2026. Verfasst von Nilesh Parihar.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid integration of Machine Learning (ML) and Deep Learning (DL) models into cybersecurity infrastructure has introduced a dual-edged paradigm: while these technologies offer unprecedented capabilities in threat detection, anomaly identification, and automated response, they simultaneously expose critical vulnerabilities to adversarial manipulation. Adversarial Machine Learning (AML) represents an emerging and sophisticated threat vector wherein malicious actors craft specially perturbed inputs — adversarial examples — designed to deceive, mislead, or bypass AI-driven security systems. This paper presents a comprehensive study and a novel multi-layered defensive framework, termed CyberShield-DNN, that addresses adversarial threats in cybersecurity contexts. We systematically investigate prominent adversarial attack methodologies including Fast Gradient Sign Method (FGSM), Carlini-Wagner (C&W), Projected Gradient Descent (PGD), and GAN-based adversarial sample generation. Our proposed architecture integrates adversarial training, input preprocessing, ensemble-based detection, and explainability modules using SHAP (SHapley Additive exPlanations) to ensure transparent decision-making. Experimental evaluation across four benchmark datasets — CICIDS-2022, NSL-KDD 2023, UNSW-NB15 updated, and a custom university exam network dataset — demonstrates that CyberShield-DNN achieves an average detection accuracy of 97.84%, with adversarial robustness improving by 34.6% over baseline models. The framework demonstrates applicability in intrusion detection, malware classification, and phishing identification systems. This research contributes a significant advancement to the field of AI-driven cybersecurity and lays foundational groundwork for PhD-level investigation into adversarial resilience of next-generation security architectures.

Zitieren

APA

Parihar, N. (2026). Adversarial Machine Learning in Cybersecurity: A Comprehensive Framework for Threat Detection and Mitigation Using Deep Neural Networks. Dandao Xuebao/Journal of Ballistics, 38(1), 391-399. https://doi.org/10.52783/dxjb.v38.288

Vancouver

Parihar N. Adversarial Machine Learning in Cybersecurity: A Comprehensive Framework for Threat Detection and Mitigation Using Deep Neural Networks. Dandao Xuebao/Journal of Ballistics. 2026;38(1):391-399. doi:10.52783/dxjb.v38.288

BibTeX
@article{Parihar2026,
  author = {Parihar, Nilesh},
  title = {Adversarial Machine Learning in Cybersecurity: A Comprehensive Framework for Threat Detection and Mitigation Using Deep Neural Networks},
  journal = {Dandao Xuebao/Journal of Ballistics},
  year = {2026},
  volume = {38},
  number = {1},
  pages = {391--399},
  doi = {10.52783/dxjb.v38.288}
}
Kostenlos starten →