Fortifying Smart City IoT Networks: A Deep Learning-Based Attack Detection Framework with Optimized Feature Selection Using MGS-ROA

Abstract

With its rapid evolution, Internet of Things (IoT) technology has gone from connecting individual devices to enabling smart cities and widespread deployments in a wide range of businesses throughout the world. However, vulnerabilities and possible breaches in IoT networks emerge as a result of diverse devices employing different protocols and having limited processing capabilities. In this study, to use a deep learning procedure to examine the collected network traffic dataset in an Internet of Things setting and identify network assaults. The Google Colaboratory (Colab) environment is utilised to conduct this investigation utilising PySpark with Apache Spark. The research makes use of the Scikit-Learn and Keras libraries. In order to train besides evaluate the classical, the ‘CICIoT2023’ dataset is utilised. To make sure that important features are included in the testing, the datasets are reduced using the Modified Gear besides Steering-based Rider Optimization Algorithm (MGS-ROA). By utilizing the Bidirectional Gated Recurrent Unit (BiGRU), a deep learning algorithm is produced. There was a comparison between the created approach and algorithms that use machine learning and deep learning. F1 parameters, recall, accuracy, and precision were used to assess the representation’s presentation. The suggested ensemble method demonstrates outstanding performance in extensive experiments and comparisons, offering a strong answer to strengthen IoT networks.