A stacked ensemble machine learning framework designed for Internet of Things security achieved up to 99.41 percent accuracy in identifying network intrusions across benchmark datasets, according to research published in the Journal of Cloud Computing. Developed to protect connected devices against complex cyber threats, the system, titled the Stacking Ensemble-based IoT Attack Detection Model, or SE-IADM, integrates multiple classification algorithms with hybrid feature selection and data balancing. In benchmark evaluations, the architecture demonstrated strong detection capabilities in both binary classification scenarios, distinguishing normal traffic from attacks, and multiclass classification scenarios, identifying specific categories of network threats across connected environments.
The investigation addresses security challenges arising from the rapid expansion of the Internet of Things, an environment where large numbers of interconnected devices continuously exchange massive volumes of data. Conventional intrusion detection systems frequently struggle to handle the complexity, heterogeneity, and dynamic nature of modern IoT attacks. Although machine learning methods have delivered encouraging outcomes in identifying suspicious trends and malicious behavior, individual detection models often fall short when addressing the full complexity and diversity of network threats. This limitation highlights the need for an adaptive and robust detection framework capable of recognizing varied threat signatures while actively minimizing false negative rates.
To overcome the constraints of standalone algorithms, the authors designed SE-IADM as an optimized stacking ensemble uniting the capabilities of three well-known classifiers: Extreme Gradient Boosting, Decision Tree, and Adaptive Boosting. By combining these three predictive algorithms into an integrated framework, the model harnesses their collective capabilities to achieve greater classification robustness. Rather than relying on a single learning approach, the stacking ensemble synthesizes outputs across the different classifiers. The researchers noted that combining multiple diverse classifiers creates an improved model that delivers greater robustness in accuracy and enhances the overall ability to identify varied threat types.
Model optimization is supported by a hybrid feature selection method designed to identify the most relevant attributes prior to classification. The researchers employed a combination of Greedy Algorithms and Genetic Algorithms to evaluate network traffic features and isolate key data attributes. By filtering out non-essential variables, this hybrid selection technique enhances detection precision and provides increased operational flexibility, enabling the framework to adapt effectively to different attack configurations without sacrificing computational performance or model reliability when assessing complex network communications.
To resolve the widespread issue of class imbalance in network intrusion datasets, the methodology incorporates the Synthetic Minority Oversampling Technique, commonly known as SMOTE. In many Internet of Things scenarios, malicious attack instances appear far less frequently than standard network communications, which can bias classification models toward majority classes. Applying SMOTE ensures a balanced representation of minority attack classes within the dataset. According to the study, this data balancing step leads directly to improved model stability and superior detection performance across both binary and multiclass classification tasks.
The research team assessed the performance of the SE-IADM architecture using two recognized benchmarks: the University of New South Wales Network Behaviour 15 dataset and the ToNIoT dataset. In experimental evaluations using the UNSW-NB15 records, the model achieved 95.07 percent accuracy in binary classification and 82.78 percent accuracy in multiclass classification. When tested on the ToNIoT benchmark data, the ensemble achieved 99.41 percent accuracy for binary attack detection and 96.64 percent accuracy for multiclass classification, demonstrating consistent detection capabilities across distinct network environments.
The findings indicate that stacking multiple classifiers alongside hybrid feature selection and data balancing provides an effective method for detecting cyber threats in Internet of Things applications. By merging Extreme Gradient Boosting, Decision Tree, and Adaptive Boosting algorithms, the architecture increases accuracy and provides heightened flexibility for identifying diverse attack patterns. The authors conclude that the resulting SE-IADM framework represents a practical, robust solution that can be readily deployed in the field to protect connected systems against evolving security threats.
Subject of Research: Intrusion detection in Internet of Things networks using stacked ensemble machine learning
Article Title: SE-IADM: an optimized stacked ensemble model for smart IoT attack detection using hybrid feature selection
Article References: Khan, M. Z., Bokhari, M. U., Alam, M., Husain, M. S., Khan, M. Z., Ashraf, Z., & Masoodi, F. S. (2026). SE-IADM: an optimized stacked ensemble model for smart IoT attack detection using hybrid feature selection. Journal of Cloud Computing. https://doi.org/10.1186/s13677-026-00995-2
Image Credits: AI Generated
DOI: 10.1186/s13677-026-00995-2
Keywords: Internet of Things, intrusion detection systems, SE-IADM, stacked ensemble learning, Extreme Gradient Boosting, Decision Tree, Adaptive Boosting, hybrid feature selection, Greedy Algorithm, Genetic Algorithm, SMOTE, class imbalance, UNSW-NB15, ToNIoT, binary classification, multiclass classification, cyber attack detection
Cite Scienmag News
Hailey Crawford. (November 19, 2026). Stacked Ensemble Model Improves Cyber Attack Detection in Connected IoT Systems. Scienmag. https://scienmag.com/stacked-ensemble-model-improves-cyber-attack-detection-in-connected-iot-systems/
Hailey Crawford. "Stacked Ensemble Model Improves Cyber Attack Detection in Connected IoT Systems." Scienmag, 19 November 2026, https://scienmag.com/stacked-ensemble-model-improves-cyber-attack-detection-in-connected-iot-systems/. Accessed 19 November 2026.
Hailey Crawford. "Stacked Ensemble Model Improves Cyber Attack Detection in Connected IoT Systems." Scienmag. November 19, 2026. https://scienmag.com/stacked-ensemble-model-improves-cyber-attack-detection-in-connected-iot-systems/








