BOA-Ensemble: A Bobcat Optimization-Driven Ensemble Deep Learning Framework for Intrusion Detection in IIoT Environments

An intelligent and scalable intrusion detection framework is required due to increased possibility for cyberattacks caused by development of industrial Internet of Things (IIoT) systems. In order to select features optimally, this study suggests a new deep learning-based intrusion detection system that uses an ensemble classification model that combines convolutional neural network (CNN), multilayer long short-term memory (MLSTM), and artificial neural network (ANN), as well as bobcat optimization algorithm (BOA). framework is tested using X-IIoTID dataset, which contains varied and multi-dimensional information about cyber-attacks on industrial systems. To guarantee high-quality data, preprocessing methods like encoding, normalization, and imputation were used. With BOA, to were able to drastically cut feature space in half, from 95 features to 42, which greatly reduced computational complexity and overfitting. When compared to individual models, ensemble model outperformed them all, achieving an impressive 97.92% accuracy, 97.56% F1-score, and 0.985 AUC. With an inference time of only 0.59 ms per sample, model is also practical for use in real-time scenarios. Strong classification capability was validated by confusion matrix analysis, especially when it came to discriminating between normal and attack patterns. Optimal resource utilization in IIoT security environments is guaranteed by this integrated strategy, which significantly improves detection accuracy. In terms of proactive cyber threat prevention, results confirm that BOA-Ensemble framework is effective.