GBTPL : Graph‐Constrained Generative Back‐Tracing for Anomaly Propagation Path Localization in IIoT and Industry 5.0 Cyber–Physical Systems
Internet Technology Letters · 2026-07 · 9 (4)
DOI: 10.1002/itl2.70284 ↗Kurz gefasst
Fachartikel in der Zeitschrift Internet Technology Letters, online erschienen am 04.06.2026. Verfasst von Linnan Zhu und 4 weiteren Autor:innen. Die Arbeit nennt 18 Literaturangaben.
Abstract (Originalsprache)
In IIoT and Industry 5.0 cyber–physical service systems, anom alies often propagate along hidden dependencies, making timely localization difficult under noisy and incomplete observations. We propose GBTPL, a graph‐constrained generative back‐tracing framework that infers the most plausible multi‐hop propagation path backward from observed anomalies. GBTPL encodes multivariate anomaly sequences into a latent causal representation and samples candidate paths using a hybrid masked‐Transformer and reverse‐diffusion generator that strictly respects the system dependency graph. A unified causal likelihood integrates temporal consistency, structural validity, and causal strength to select the final path. Experiments on SWaT, WADI, UAV‐Log, and Synthetic‐Graph show that GBTPL achieves consistently better path‐level localization performance and stronger robustness to noise and missing logs than CausalRCA, MicroHECL, and BALANCE, while producing concise and causally plausible chains.
Zitieren
APA
Zhu, L., Cao, M., Dong, Q., Ma, M., & Xu, Z. (2026). GBTPL : Graph‐Constrained Generative Back‐Tracing for Anomaly Propagation Path Localization in IIoT and Industry 5.0 Cyber–Physical Systems. Internet Technology Letters, 9(4). https://doi.org/10.1002/itl2.70284
Vancouver
Zhu L, Cao M, Dong Q, Ma M, Xu Z. GBTPL : Graph‐Constrained Generative Back‐Tracing for Anomaly Propagation Path Localization in IIoT and Industry 5.0 Cyber–Physical Systems. Internet Technology Letters. 2026;9(4). doi:10.1002/itl2.70284
BibTeX
@article{Zhu2026,
author = {Zhu, Linnan and Cao, Mingjing and Dong, Qingyun and Ma, Manzeng and Xu, Zhonglin},
title = {GBTPL : Graph‐Constrained Generative Back‐Tracing for Anomaly Propagation Path Localization in IIoT and Industry 5.0 Cyber–Physical Systems},
journal = {Internet Technology Letters},
year = {2026},
volume = {9},
number = {4},
doi = {10.1002/itl2.70284}
}
EUVIMED