A Generative Diffusion-Based Framework and Optimization-Driven Hybrid Deep Learning Architecture for IoT Intrusion Detection
Journal of Intelligent Decision Making and Information Science · 09.09.2026 · 3 (18s) · 3107-3120
DOI: 10.17762/jidmis.v3.6072 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 09.09.2026. Verfasst von Waleed Khan. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The rapid expansion of the Internet of Things (IoT) has increased the complexity of securing interconnected devices against distributed denial-of-service (DDoS), denial-of-service (DoS), reconnaissance, botnet, and other malicious network activities. Conventional intrusion detection systems (IDSs) often experience reduced performance when confronted with highly imbalanced traffic distributions, rare attack classes, evolving attack patterns, and high-dimensional network features. This study proposes a Generative Diffusion-Based and Optimization-Driven Hybrid Deep Learning Framework (GD-OHDF) for robust IoT intrusion detection. The proposed architecture integrates a conditional tabular diffusion model for minority-class traffic generation, a hybrid convolutional neural network and bidirectional long short-term memory network (CNN-BiLSTM) for spatial and sequential feature learning, and a metaheuristic optimization strategy for hyperparameter selection. A lightweight attention mechanism is incorporated to emphasize informative traffic features, while threshold calibration is employed to control false-positive rates. The framework is designed for evaluation on publicly available network intrusion datasets, including BoT-IoT, TON-IoT, and Edge-IIoTset. Its performance is assessed using accuracy, precision, recall, macro-F1 score, area under the receiver operating characteristic curve (ROC-AUC), false-positive rate, inference latency, and computational overhead. Five result tables and five graphical analyses are specified to examine classification performance, class imbalance, convergence, ablation effects, and cross-dataset generalization. Furthermore, the proposed framework aims to improve the identification of underrepresented attack categories while maintaining a balance between detection performance and computational efficiency. Comparative experiments with conventional machine learning models and alternative data augmentation techniques will be used to evaluate the contribution of each architectural component. The study also considers the challenges of data leakage, synthetic sample quality, and distribution shifts between different IoT network environments. These investigations are expected to provide a systematic basis for assessing the reliability, scalability, and practical applicability of diffusion-assisted intrusion detection. The framework provides a reproducible research direction for integrating generative augmentation, hybrid deep learning, and optimization in resource-constrained IoT security environments.
Zitieren
APA
Khan, W. (2026). A Generative Diffusion-Based Framework and Optimization-Driven Hybrid Deep Learning Architecture for IoT Intrusion Detection. Journal of Intelligent Decision Making and Information Science, 3(18s), 3107-3120. https://doi.org/10.17762/jidmis.v3.6072
Vancouver
Khan W. A Generative Diffusion-Based Framework and Optimization-Driven Hybrid Deep Learning Architecture for IoT Intrusion Detection. Journal of Intelligent Decision Making and Information Science. 2026;3(18s):3107-3120. doi:10.17762/jidmis.v3.6072
BibTeX
@article{Khan2026,
author = {Khan, Waleed},
title = {A Generative Diffusion-Based Framework and Optimization-Driven Hybrid Deep Learning Architecture for IoT Intrusion Detection},
journal = {Journal of Intelligent Decision Making and Information Science},
year = {2026},
volume = {3},
number = {18s},
pages = {3107--3120},
doi = {10.17762/jidmis.v3.6072}
}
EUVIMED