Adaptive Feature Selection and Deep Learning Synergy for IoT Security and Energy Prediction Using Tree Growth and Rooster Optimization

Abstract

The rapid evolution of the Internet of Things (IoT) ecosystem has introduced new dimensions of intelligent connectivity, enabling seamless integration between smart devices and critical energy systems. However, this advancement has simultaneously expanded the attack surface, exposing IoT networks to security breaches, intrusions, and data manipulation that threaten both privacy and energy management systems. This paper presents an innovative hybrid framework that combines adaptive feature selection with deep learning models, optimized through Tree Growth Optimization (TGO) and Rooster Optimization Algorithm (ROA). The proposed model efficiently identifies relevant features from heterogeneous IoT data streams, enhancing detection precision and forecasting accuracy. TGO facilitates structured feature extraction by modeling environmental adaptation, while ROA ensures global optimization by simulating rooster hierarchy and dominance behavior. The synergy between these algorithms optimizes a deep learning architecture comprising a hybrid LSTM-DBN network for intrusion detection and renewable energy forecasting. Experimental analysis on benchmark IoT and energy datasets demonstrates that the proposed hybrid model significantly outperforms conventional optimization and learning approaches in accuracy, computational efficiency, and robustness. Results confirm that this adaptive synergy framework can serve as a foundation for secure, sustainable, and intelligent IoT infrastructures.