EUVIMEDEUVIMED entdecken
Crossref

Digital Twins Reimagined: Zero-Day LLM-Powered Moving Target Defense In-Depth for Real-Time CPS

Rajarshi Mukherjee, Mohamed Azab, Thidapat Chantem

IEEE Internet of Things Journal · 15.06.2026 · 13 (12) · 27001-27012

DOI: 10.1109/jiot.2026.3679937 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Internet of Things Journal, erschienen am 15.06.2026. Verfasst von Rajarshi Mukherjee und 2 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Gefördert von: National Science Foundation, Commonwealth Cyber Initiative, an Investment in the Advancement of Cyber Research and Development, Innovation, and Workforce Development.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The Internet of Things (IoT) revolution has profoundly transformed modern industries by embedding connected digital devices across critical sectors such as energy, pharmaceuticals, and manufacturing. Such interconnected devices together form cyber-physical systems (CPSs), complex structures that enable computation to directly control physical processes. These systems increasingly rely on automation and real-time control to enhance efficiency, productivity, and safety. However, this growing dependence on connectivity has significantly expanded the attack surface, exposing mission-critical infrastructure to sophisticated cyber threats. Among the most dangerous are zero-day vulnerabilities, which are unknown flaws in software or firmware, and zero-day design flaws, which are deeply embedded architectural weaknesses often undetectable by conventional verification methods. Traditional defense mechanisms, largely based on static rules or known threat signatures, are ill-equipped to detect these stealthy and evolving threats—especially in real-time systems, where even minor delays in detection or response can result in catastrophic outcomes. To overcome these limitations, this work presents a transformative security framework that leverages the reasoning and generative capabilities of large language models (LLMs) to construct an intelligent, redundant digital twin of the operational system. Unlike traditional replicas, the LLM-driven digital twin is logically and architecturally isolated, making it inaccessible to attackers targeting the primary system. Crucially, the twin is not a static copy: it is dynamically synthesized and continuously restructured by the LLM to ensure a heterogeneous and evolving codebase. This ensures functional alignment with the physical system while introducing deliberate software diversity, thereby increasing attacker uncertainty and resilience to shared exploits. The central hypothesis of the framework is that semantic deviations between the decisions or behaviors of the digital twin and its physical counterpart can serve as reliable indicators of anomalous activity—whether stemming from external attacks or latent design flaws. By monitoring this divergence, the system facilitates the early detection of zero-day threats and supports real-time mitigation strategies, thereby significantly strengthening the resilience of critical infrastructure. Implementation results demonstrated the effectiveness and efficiency of the proposed approach.

Förderung

  • National Science Foundation
  • Commonwealth Cyber Initiative, an Investment in the Advancement of Cyber Research and Development, Innovation, and Workforce Development

Zitieren

APA

Mukherjee, R., Azab, M., & Chantem, T. (2026). Digital Twins Reimagined: Zero-Day LLM-Powered Moving Target Defense In-Depth for Real-Time CPS. IEEE Internet of Things Journal, 13(12), 27001-27012. https://doi.org/10.1109/jiot.2026.3679937

Vancouver

Mukherjee R, Azab M, Chantem T. Digital Twins Reimagined: Zero-Day LLM-Powered Moving Target Defense In-Depth for Real-Time CPS. IEEE Internet of Things Journal. 2026;13(12):27001-27012. doi:10.1109/jiot.2026.3679937

BibTeX
@article{Mukherjee2026,
  author = {Mukherjee, Rajarshi and Azab, Mohamed and Chantem, Thidapat},
  title = {Digital Twins Reimagined: Zero-Day LLM-Powered Moving Target Defense In-Depth for Real-Time CPS},
  journal = {IEEE Internet of Things Journal},
  year = {2026},
  volume = {13},
  number = {12},
  pages = {27001--27012},
  doi = {10.1109/jiot.2026.3679937}
}
Kostenlos starten →