An Intelligent Deep Learning Framework for Anomaly Detection in IoT Cybersecurity

Abstract

Network traffic anomalies or intrusion should be revealed in real-time as a result of the higher frequency of cyberattacks and the high frequency of network traffic. The reason for the ineffectiveness of manual inspection is the significant volume, speed, and variability of network traffic data. The proposed research proposes an effective cybersecurity anomaly detection model of the Internet of Things (IoT) that is founded on an integration of ANN and LSTM models. It used IoT-23 dataset which comprised of multiple malware and normal traffic captures, followed by the preprocessing phase of cleaning, normalizing, and feature selection. The goal of all exploratory data analysis and preprocessing, including feature selection and normalization, is to enhance data quality and model performance. With a 98.6% accuracy, precision, and F1-score of 98, experimental data show that the Hybrid ANN+LSTM model outperforms more conventional ML models like ANN, SVM, and Naive Bayes. These results give strength to the concept that the proposed framework could be a viable solution to next-gen IoT security systems, as it enables effective anomaly detection even in highly dynamic IoT environments.