Cyberattack Detection in Smart Urban Networks Using Attention-Βased Bi-Directional Recurrent Neural Networks and the Nature-Inspired Mellivora Heron Optimization

Abstract

The smart city concept combines Information and Communication Technology (ICT) and Internet of Things (IoT) from physical (actuators and sensors) and non-physical (e.g., exterior databases) data sources to establish services. This paper presents an Intelligent Cybersecurity Framework for Smart City Networks Using the Mellivora Heron Optimization Algorithm (ICFSCN-MHOA). The primary focus of ICFSCN-MHOA is to provide a scalable and effective solution for detecting and mitigating cyber threats in real-time smart urban environments. Initially, data normalization is performed using z-score normalization. The Fox Optimizer Algorithm (FOA) is used to detect and select the most significant features to reduce dimensionality and eliminate redundancy. A hybrid Bidirectional Gated Recurrent Unit with a Channel Attention Mechanism (BiGRU-CAM) is utilized to classify cyberattacks in smart city networks. The hyperparameter fine-tuning process of the hybrid BiGRU-CAM model is performed using the Mellivora Heron Optimization Algorithm (MHOA). To validate the improved predictive results of ICFSCN-MHOA, a thorough study was conducted using the CIC-IDS2017 dataset, where the model achieved a superior accuracy of 99.19% over recent approaches.