Improvising Cybersecurity in IoT Networks Using Machine Learning for Intrusion Identification

Abstract

Smart cities, industrial automation, healthcare, and the Internet of Things have all been profoundly affected by the exponential expansion of the IoT, which has enabled massive connection and intelligent decision-making. The ever-changing, varied, and sometimes under-resourced nature of IoT devices makes intrusion detection all the more important in protecting these networks from cyberattacks. The drawbacks of classic intrusion detection systems (IDS) include issues with scalability, high false positive rates, and the inability to identify sophisticated attacks such as zero-day vulnerabilities. To get over these problems, this study presents a robust intrusion detection model that optimizes features and leverages deep learning. To ensure high-quality input utilizing the UNSW-NB15 dataset, preprocessing procedures for data were utilized, such as normalization, resolving missing values, and balancing with SMOTE. Employed the Butterfly Optimization Algorithm (BOA) to boost computing efficiency and cut down on redundant features. Then, a Deep Neural Network (DNN) was trained with F1-score, recall, accuracy, and precision exceeding 98.04%. The suggested framework's better performance and scalability are brought to light through comparative study with existing frameworks. Lightweight adaptations for IoT devices with limited resources, real-time deployment, and evaluation across datasets the main areas of future development.