CNS-STGU Defense: Trust-Conditioned Cross-Dataset IoT Intrusion Detection with Unknown-Family Rejection and Drift-Aware Response Control
Journal of Information and Communications Technology: Algorithms, Systems And Applications · 2026
DOI: 10.64189/ict.26317 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Information and Communications Technology: Algorithms, Systems And Applications, 2026. Verfasst von Atheer Alaa Hammad und 2 weiteren Autor:innen.
Abstract (Originalsprache)
In the case of Internet of Things (IoT) intrusion detection, traffic classification is a necessary step, as well as a decision on whether there is enough evidence to warrant a response. This study introduces a trust-conditioned spatial–temporal framework, called CNS-STGU Defense, which splits these tasks. A typical representation is to map CICIoT2023 and TON_IoT Network to 6 behavioural families and 33 packet-derived descriptors, where the grouping of capture/device/session is done before fitted preprocessing and window construction. The encoder is defined by a one-dimensional convolutional neural network (CNN), bidirectional long short-term memory (BiLSTM) network and additive attention. Decision acceptance is based on calibrated probabilities, prototype distance, uncertainty and distribution monitoring, and simulator-ranked actions are limited by symbolic constraints. The provided 187,360-window reference matrix achieves 99.495% accuracy, macro-averaged F1 score (macro-F1) of 99.505%, Matthew’s correlation coefficient (MCC) of 0.9939 and false-alarm rate (FAR) of 0.603%. The macro-F1 for balanced transfer reconstruction is 94.207% and 94.707%. Evidence is broken down into matrix derived quantities (M), deterministic analytical calculations (A), and reconstructed or stated illustrations (R). The contribution consists of an auditable interface connecting prediction, acceptance and response constraints, numerical consistency analysis and a matched evaluation protocol. The illustrations of the transfer and gate do not constitute an empirical superiority. Retained predictions, fitting records and response logs are not provided here and must be validated.
Zitieren
APA
Hammad, A. A., Oza, K. S., & Kharade, K. G. (2026). CNS-STGU Defense: Trust-Conditioned Cross-Dataset IoT Intrusion Detection with Unknown-Family Rejection and Drift-Aware Response Control. Journal of Information and Communications Technology: Algorithms, Systems And Applications. https://doi.org/10.64189/ict.26317
Vancouver
Hammad AA, Oza KS, Kharade KG. CNS-STGU Defense: Trust-Conditioned Cross-Dataset IoT Intrusion Detection with Unknown-Family Rejection and Drift-Aware Response Control. Journal of Information and Communications Technology: Algorithms, Systems And Applications. 2026. doi:10.64189/ict.26317
BibTeX
@article{Hammad2026,
author = {Hammad, Atheer Alaa and Oza, Kavita S. and Kharade, Kabir G.},
title = {CNS-STGU Defense: Trust-Conditioned Cross-Dataset IoT Intrusion Detection with Unknown-Family Rejection and Drift-Aware Response Control},
journal = {Journal of Information and Communications Technology: Algorithms, Systems And Applications},
year = {2026},
doi = {10.64189/ict.26317}
}
EUVIMED