A Hybrid Intelligent Model for Spam Detection in IoT Communication Networks

Abstract

The Internet of Things (IoT) has rapidly expanded into a global ecosystem of interconnected devices, enabling automation and intelligent decision-making across domains such as healthcare, smart homes, and industrial systems. However, this growth has also amplified security risks, with spam attacks manifesting as fake sensor readings, unauthorized commands, or traffic flooding posing severe threats to device reliability and data integrity. To counter these challenges, this study proposes an intelligent spam detection framework that integrates edge-level monitoring, IoT-specific feature engineering and machine learning (ML) techniques. The framework leverages statistical traffic features and advanced classifiers to differentiate between benign and malicious activity, ensuring dynamic, real-time detection suitable for deployment in resource-constrained IoT environments. Experimental validation was carried out using an adapted version of the REFIT Smart Home dataset, simulating both legitimate and adversarial traffic. Five ML models Logistic Regression, Support Vector Machines, Random Forest, Gradient Boosting, and Neural Networks were comparatively evaluated using accuracy, precision, recall, F1-score, and ROC-AUC as performance metrics. Results showed that Neural Networks achieved the highest accuracy (95.1%) and recall (95.7%), outperforming other models in capturing complex spam patterns. Random Forest and Gradient Boosting also demonstrated strong reliability, while Logistic Regression offered a lightweight, resource-efficient option. These findings affirm that machine learning, particularly deep and ensemble models, provides a robust pathway to securing IoT ecosystems against evolving spam threats.