EUVIMEDEUVIMED entdecken
Crossref

Scalable Privacy-Preserving Cyber Defense: Federated Self-Supervised Learning for Zero-Day Threat Detection in Critical Infrastructure

Sadiya Afrin, Jawad Sarwar

International Journal of Innovative Science and Research Technology · 07.05.2026 · 10

DOI: 10.38124/ijisrt/26may248 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Innovative Science and Research Technology, online erschienen am 07.05.2026. Verfasst von Sadiya Afrin und 1 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The high rate of digitalization of major infrastructural systems, such as energy grids, transportation systems, healthcare services, and industrial control systems, has greatly exposed them to advanced cyberattacks. Of these threats, zero-day attacks are especially dangerous because they have an unknown signature and cannot be detected by traditional signature-based detection mechanisms before they cause any damage. Simultaneously, centralized security analytics solutions also come with significant privacy, regulatory, and operational risks, since sensitive operational data cannot be easily distributed across distributed facilities. The proposed study suggests a Federated Self-Supervised Federated Defense framework that combines both federated learning and self-supervised repair learning to allow privacy-preserving collaborative zero-day threat detection across geographically distributed network nodes of infrastructure. This is made possible by the suggested architecture, which enables local systems to utilize models trained on-site, while only sharing some encrypted model parameters, ensuring confidentiality is maintained and regulatory compliance is upheld. Self-supervised pretraining is more sensitive to anomalies by learning inherent behavioral patterns based on unlabeled network and system telemetry, which is more useful for generalization to unseen attacks. Experimental analysis with simulated infrastructure datasets that are distributed shows that the detection accuracy is higher, the number of false positives is lower, and the communication overhead is less than with centralized and purely supervised baselines. The framework is also resistant to data heterogeneity and adversarial manipulation via secure aggregation and adaptive model updates. Findings indicate that federated self-supervised learning can be utilized to substantially enhance collective cyber defense without compromising privacy or operational independence. This study highlights a scalable and reliable future for next-generation smart surveillance of distributed critical infrastructure sites.

Zitieren

APA

Afrin, S., & Sarwar, J. (2026). Scalable Privacy-Preserving Cyber Defense: Federated Self-Supervised Learning for Zero-Day Threat Detection in Critical Infrastructure. International Journal of Innovative Science and Research Technology, 10. https://doi.org/10.38124/ijisrt/26may248

Vancouver

Afrin S, Sarwar J. Scalable Privacy-Preserving Cyber Defense: Federated Self-Supervised Learning for Zero-Day Threat Detection in Critical Infrastructure. International Journal of Innovative Science and Research Technology. 2026:10. doi:10.38124/ijisrt/26may248

BibTeX
@article{Afrin2026,
  author = {Afrin, Sadiya and Sarwar, Jawad},
  title = {Scalable Privacy-Preserving Cyber Defense: Federated Self-Supervised Learning for Zero-Day Threat Detection in Critical Infrastructure},
  journal = {International Journal of Innovative Science and Research Technology},
  year = {2026},
  pages = {10},
  doi = {10.38124/ijisrt/26may248}
}
Kostenlos starten →