Navigating the IoT Landscape: A Deep Dive into Anomaly Classification for IoT Security
Jisha Jose, Judith J.E. · International Journal of Electronics and Communication Engineering · 2024
The Internet of Things (IoT) is crucial for technological advancement, attracting significant interest from researchers worldwide. However, the exponential growth of IoT devices and their huge volumes of data introduce substantial risks related to several security threats and vulnerabilities. The increasing implementation of IoT infrastructure has led to challenges such as device failures, elevated risks, and greater exposure to attacks, anomalies, and potential security breaches. Tackling and alleviating these concerns represent a critical area of focus within the broader field of IoT. By utilizing the IoTID20 dataset, precisely designed for IoT anomaly detection, the study suggests a novel approach for anomaly classification in IoT environments using a Deep Learning (DL) model optimized by a Normalized Bayesian Optimization Algorithm (NBOA) and a Convolutional Neural Network (CNN) architecture is employed for classifying anomaly, benefiting from the spatial pattern recognition capabilities of CNNs. The study employs a Machine Learning (ML) approach that utilizes Decision Tree (DT) and feature selection through the Harris Hawk Optimization (HHO) algorithm. The study validates the efficiency of using NBOA and HHO in boosting anomaly classification, ensuring faster convergence and improved accuracy. The outcomes demonstrate the superior performance of the DL model with 95.67% accuracy, surpassing the proposed ML and other state-of-the-art models. Combining these advanced optimization techniques with the DL and ML models, the study addresses the security challenges in the rapidly expanding IoT landscape, offering a robust solution for real-time anomaly detection.