The Internet of Things (IoT) is a global network of uniquely capable of being addressed, networked items that use sensory features, protocol for communication, computational resources, as well as information analysis capabilities to offer services that interpret data. The innovative security system Lightweight Intelligent Optimized Detector (LIO-Det) has been designed to defend the security and privacy of cloud IoT systems against modern cyberattacks. A single hot encoding approach is used for both data preprocessing and cleaning, which improves the classifier's efficiency. In addition, the Upgraded Butterfly—Lagrange Interpolated Optimization (UBLIO) technique is used to select the most important features with higher search efficiency from the provided data. Moreover, the innovative Weighted Stacking Learning Classifier (WSLC) is used to classify attacking and normal events with a high performance rate and accuracy of the system. Performance is assessed in this study using the most popular and recent cyberattack datasets, such as IoT-POT, MQTT, CICIDS 2017, and UNSW-NB 15. Furthermore, the results show that the suggested LIO-Det model performs better than other current methods, with a 99% accuracy gain and a 1.5% false detection rate decrease.