Machine Learning and Deep Learning Approaches for Cyberattack Detection in IoT Environments

Abstract

Intrusion detection has become a crucial field of study due to the quick expansion of Internet of Things (IoT) devices and the growing complexity of cyber threats. With an emphasis on their evaluation settings, advantages, and disadvantages, this paper examines current Machine Learning (ML) and Deep Learning (DL) techniques for cyberattack detection in Internet of Things contexts. Current research frequently relies on a small number of datasets, which limit the capacity to evaluate model generalization and robustness in a variety of contexts. This study uses five benchmark datasets UNSW_NB15, NSL_KDD, Edge-IIoTset, RT-IoT2022, and IoTID20 to compare ML and DL models in order to address this problem. Additionally, the Manta Ray Foraging Optimizer (MRFO) and other metaheuristic optimization techniques are emphasized as a means of enhancing model performance through ensemble learning and hyperparameter adjustment. Two optimized models, ET-MRFO and VC-RXE-MRFO, are analyzed as examples of enhanced detection accuracy and stability. The results indicate that optimized ML ensemble models achieve competitive performance compared to DL methods.