The exponential advance of the Internet of Things (IoT) has given rise to the epoch of pervasive connectivity, with billions of devices gathering, processing, and transmitting data through numerous areas. This interconnected network, however, offers an intricate threat environment. The limited processing power and memory of many IoT devices make them prime targets for cyber threats. Conventional security solutions regularly verify insufficient in this dynamic and ever-changing security landscape. Artificial intelligence (AI), with its potential for intellectual attack detection, response, and anomaly classification, has developed as an evolutionary force in safeguarding IoT networks. Instead, cyber attackers have thought about developing AI and utilizing adversarial AI to perform cyber-security attacks. Therefore, this study presents a novel Random Vector Functional Link-Enhanced Metaheuristic Algorithms for Robust Cyber Threat Mitigation (RVFLEMA-RCTM) technique. The aim is to automate the recognition and classification of cyber-attacks in IoT sensor networks. Initially, The RVFLEMA-RCTM technique applies min–max normalization to standardize raw data. The adaptive grey wolf optimizer (AGWO) technique is used to derive feature subsets for feature subset selection. Besides, the random vector functional link (RVFL) model is employed for the automated detection of cyberattacks in IoT. Finally, an improved dung beetle optimizer (IDBO) model is used to optimize the parameter tuning of the RVFL model. The simulation study of the RVFLEMA-RCTM model is performed under several measures. The performance validation of the RVFLEMA-RCTM method portrayed a superior accuracy value of 99.33% over existing methodologies.