<![CDATA[General Background: The rapid expansion of Internet of Things networks has intensified cybersecurity challenges, particularly distributed denial-of-service attacks targeting interconnected devices. Specific Background: Intrusion Detection Systems based on machine learning often suffer from high computational complexity and reduced performance when processing large-scale datasets due to manual feature extraction and dimensionality issues. Knowledge Gap: Existing approaches lack efficient hybrid frameworks that integrate advanced optimization algorithms with ensemble classification to address large modern intrusion datasets such as NSL-KDD and UNSW-NB15. Aims: This study proposes an Intrusion Detection System integrating the Enhanced Flower Pollination Algorithm for optimal feature selection with an ensemble classification framework combining Random Forest, Decision Tree, and Support Vector Machine. Results: Experimental evaluation achieved accuracy rates of 99.67% on NSL-KDD and 99.32% on UNSW-NB15, demonstrating reduced computational complexity and improved detection capability across multiple attack categories. Novelty: The study introduces a hybrid EFPA-based feature selection strategy integrated with majority voting ensemble classification for IoT security environments. Implications: The proposed framework supports scalable, high-accuracy intrusion detection suitable for real-time IoT deployments and provides a foundation for future integration with advanced security infrastructures. Keywords: Intrusion Detection System, Internet of Things, Enhanced Flower Pollination Algorithm, Ensemble Classification, Network Security Key Findings Highlights: Hybrid optimization reduced dimensionality while preserving critical attack indicators. Majority voting integration increased model generalization across attack categories. High detection performance achieved on two benchmark cybersecurity datasets.]]>
