<p>Internet of Things (IoT) technology has evolved beyond personal devices to power global deployments across a wide range of networks which has a significant impact on global commerce. However, security challenges arise due to the wide range of protocols and computational capabilities in IoT devices. To combat these issues, a novel hybrid optimization-enabled neural network for classification of intrusion data against IoT system (Hybrid-CID) is proposed particularly to identify intrusions in resource-constrained IoT devices. Initially, the incoming data are standardized by removing the irrelevant information through preprocessing to ensure the performance of detection models. After preprocessing, the Hybrid-CID framework develops a hybrid optimization algorithm to identify the intrusions from the traffic data which ensures data privacy by maintaining the reliability and integrity of IoT deployments. Finally, the combined deep learning (DL) network classifies the identified intrusions to contribute proactive threat mitigation by ensuring the confidentiality of IoT system data. The Hybrid-CID system is validated through the benchmark CSE-CIC-IDS 2018 and CICIDS 2017 datasets using accuracy, specificity, precision, F1-score, recall, execution time, communication cost, detection rate, detection time, and computational cost. The Hybrid-CID framework achieves an overall accuracy of 97.82%, whereas the WDLSTM, TLBO-IDS, and DIS-IoT techniques achieve 87.42%, 89.58%, and 94.72%, respectively, for efficiently detecting intrusions in IoT networks.</p>