A Secure and Privacy-Preserving IoT Cybersecurity Framework Using Feature Selection and Ensemble Deep Learning for Smart City Applications

The rapid proliferation of Internet of Things (IoT) applications in recent years has played a significant role in the advancement of smart cities. Sustainable cities and communities are a main goal of the Sustainable Development Goal (SDG), which aims to make urban areas safe, resilient, and sustainable by 2030. Smart cities employ IoT-based technologies, communication systems, and intelligent applications to optimize operational efficiency and improve both service delivery and citizens’ quality of life. Cybersecurity has become a problematic issue in IoT environments, needing effective addressing of the persistent cyberthreats. Intrusion Detection Systems (IDSs) are required to protect data, and the utilization of Artificial Intelligence (AI) subfields such as Machine Learning (ML) and Deep Learning (DL) is proven to be among the most effective solutions. In this paper, an Artificial Intelligence-Driven Cybersecurity Detection using Ensemble Models (AIDCD-EM) framework in smart city applications is proposed. Initially, the Z-score normalization is used for data normalization. For dimensionality reduction, the AIDCD-EM utilizes the enhanced Mutual Information Feature Selection (MIFS) method and an ensemble classification comprising the Bi-directional Gated Recurrent Unit (BiGRU), autoencoder (AE), and Graph Convolutional Networks (GCNs), is utilized for cyberattack classification. The experimental valuation of the AIDCD-EM model highlighted superior accuracy values of 99.60% and 99.55% when investigated under the ToN-IoT and Edge-IIoT datasets.