<scp>ResA</scp> ‐ <scp> D ² PySepCo </scp> : Cyber‐Attack Detection in Fog Based <scp>IoT</scp> Network Using Pyramidal Dilated Separable Residual Convolutional With Effective Optimization Algorithm

ABSTRACT Cyberattack detection systems are critical in protecting the Internet of Things (IoT) network from attacks. Cyber‐attack detection is critical in fog‐based IoT systems to protect against a wide range of threats, such as illegal access, data breaches, and service disruptions. Present cyber‐attack detection models have limitations, such as a high false alarm rate (FAR) and limited detection accuracy. Therefore, it is essential to create reliable cyberattack detection systems that can raise the accuracy of detection and lower the rate of false alarms. This paper presents a deep learning (DL)‐based classifier model with an effective feature extraction mechanism for detecting cyber‐attacks in a fog‐based IoT system. Before initiating the process, data augmentation has been done to balance the dataset. Initially, the input data are collected from the public source dataset; the collected data are passed into the pre‐processing stage to rescale the data properly, which is performed by Z ‐Score normalization. The optimal set of features from the pre‐processed data is extracted by a meta‐heuristic technique, namely the Hybrid white shark sea lion optimization algorithm (Hy‐WS 2 LO). After selecting optimal features, the data is fed into a classifier model to detect cyber‐attacks in a fog‐based IoT environment. An effective Residual attention‐based dilated pyramidal depth‐wise separable convolution (ResA‐D 2 PySepCo) is used for identifying cyber‐attacks in fog‐based IoT networks. The proposed model can obtain an accuracy of 98.87% and 99.74% for ToN IoT and CICIDS 2018 datasets.