EUVIMEDEUVIMED entdecken
Studienprotokoll

PREMADC: Protocol Reverse Engineering via Multiagent Analysis and Deep Clustering for Industrial Control Protocols

Xuejun Zong, Xinxu Gao, Dong Li, Kan He, Lian Lian, Hongyan Shi, Yifei Sun, Bowei Ning

IEEE Internet of Things Journal · 15.07.2026 · 13 (14) · 31240-31254

DOI: 10.1109/jiot.2026.3685605 ↗

Kurz gefasst

Studienprotokoll in der Zeitschrift IEEE Internet of Things Journal, erschienen am 15.07.2026. Verfasst von Xuejun Zong und 7 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben. Gefördert von: Major Science and Technology Project of Liaoning Province of China, Applied Basic Research Program of Liaoning Province, Natural Science Foundation of Liaoning Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

With the development of the industrial Internet of Things (IIoT), nonstandard and proprietary industrial control protocols (ICPs) are widely deployed, posing significant challenges to protocol understanding and security analysis. Protocol reverse engineering (PRE), which aims to analyze unknown protocols from observed traffic under plaintext or partially encrypted payloads, is a fundamental prerequisite for network security research. However, existing semantic inference methods still depend heavily on tedious manual inspection, resulting in long analysis cycles and poor scalability. To address these limitations, we propose PREMADC. It leverages unsupervised SimCSE to learn protocol embeddings and introduces a probabilistic-boundary-based soft silhouette clustering loss to enhance embedding discriminability and intercluster separation. On top of the clustering results, PREMADC further builds a progressive multiagent system with three specialized agents for parsing, comparison, and modeling, enabling structured and verifiable semantic inference from passive traffic. Given the multiagent reasoning overhead, PREMADC is designed for offline or quasi-online reverse engineering. Experiments on a typical energy-enterprise industrial attack–defense platform show that PREMADC achieves an average semantic inference accuracy that significantly outperforms representative baselines, including Netzob, NetPlier, and BinaryInferno. Moreover, a case study on GE-SRTP, an under-documented proprietary protocol, indicates PREMADC’s capability to identify key fields in truly unknown protocols.

Förderung

  • Major Science and Technology Project of Liaoning Province of China
  • Applied Basic Research Program of Liaoning Province
  • Natural Science Foundation of Liaoning Province
  • Science and Technology Program of Liaoning Province
  • Liaoning Provincial Science and Technology Innovation Platform Program

Zitieren

APA

Zong, X., Gao, X., Li, D., He, K., Lian, L., Shi, H., Sun, Y., & Ning, B. (2026). PREMADC: Protocol Reverse Engineering via Multiagent Analysis and Deep Clustering for Industrial Control Protocols. IEEE Internet of Things Journal, 13(14), 31240-31254. https://doi.org/10.1109/jiot.2026.3685605

Vancouver

Zong X, Gao X, Li D, He K, Lian L, Shi H, et al. PREMADC: Protocol Reverse Engineering via Multiagent Analysis and Deep Clustering for Industrial Control Protocols. IEEE Internet of Things Journal. 2026;13(14):31240-31254. doi:10.1109/jiot.2026.3685605

BibTeX
@article{Zong2026,
  author = {Zong, Xuejun and Gao, Xinxu and Li, Dong and He, Kan and Lian, Lian and Shi, Hongyan and Sun, Yifei and Ning, Bowei},
  title = {PREMADC: Protocol Reverse Engineering via Multiagent Analysis and Deep Clustering for Industrial Control Protocols},
  journal = {IEEE Internet of Things Journal},
  year = {2026},
  volume = {13},
  number = {14},
  pages = {31240--31254},
  doi = {10.1109/jiot.2026.3685605}
}
Kostenlos starten →