Hybrid deep learning ensemble with score-based feature optimization for cyber attack detection in IoT systems

Abstract

The rapid growth of internet of things (IoT) devices have improved connectivity but also exposed networks to cyber threats. This study proposes a prediction-scoring-based ensemble deep learning model with prediction-scoring-optimized feature selection (EDLM-PSOFS) for intrusion detection in IoT systems. The model integrates random forest (RF) feature extraction with ant lion optimization (ALO)-tuned convolutional neural networks (CNNs) to balance accuracy and computational efficiency. Using the KDD Cup ’99 dataset containing 4.9 million traffic records and 41 features, the framework achieved 97% accuracy, 0.99 precision, and 0.97 recall within five epochs. Comparative evaluation shows faster convergence and reduced complexity than gated recurrent units (GRU), long short-term memory (LSTM), and support vector machine (SVM) baselines, demonstrating suitability for real-time, resource-constrained IoT deployments.