MRFO-MOENET: A MIXTURE OF EXPERTS TRANSFER LEARNING MODEL OPTIMIZED BY MANTA RAY FORAGING ALGORITHM FOR SMART IOT INTRUSION DETECTION

Abstract

With the rapid growth of the Internet of Things (IoT), ensuring the security of connected devices has become increasingly critical due to the rising frequency of sophisticated cyber-attacks. This paper presents MRFO-MoENet, a novel hybrid approach that combines deep transfer learning with a Mixture of Experts (MoE) ensemble architecture, optimized using the Manta Ray Foraging Optimization (MRFO) algorithm, to enhance IoT attack detection. The proposed framework leverages multiple pre-trained deep learning models, transferring knowledge from large-scale domains to IoT-specific data, and employs a dynamic expert selection mechanism for ensemble prediction. MRFO is utilized to fine-tune the parameters of both individual experts and the gating network, enabling optimal collaboration among models. Experiments conducted on the Edge-IIoTset dataset demonstrate that MRFO-MoENet achieves a high detection accuracy of 99.92% while maintaining a very low false alarm rate. These results confirm the robustness and practical applicability of our method for securing IoT environments against modern cyber threats.