IOT CYBER ATTACK DETECTION USING MACHINE LEARNING AND DEEP LEARNING

Abstract

The rapid proliferation of Internet of Things (IoT) devices has transformed various sectors, enhancing efficiency and connectivity. However, this expansion has also introduced significant vulnerabilities, making IoT systems attractive targets for cyberattacks. This paper presents a comprehensive approach to IoT cyber-attack detection, addressing the unique challenges posed by the diverse and often resource-constrained nature of IoT environments. By integrating machine learning and deep learning algorithms with anomaly detection techniques, the proposed system aims to identify and mitigate potential threats in real time, ensuring the integrity and security of IoT networks. The detection mechanism is designed to monitor network traffic and device behaviour continuously, learning normal operational patterns to effectively identify deviations indicative of cyber threats. The system was trained and evaluated on the CICAPT-IIoT dataset (over 4 GB, comprising 95% normal records and 5% attack records) using four algorithms: Random Forest, Support Vector Machine (SVM), Logistic Regression, and a 2D Convolutional Neural Network (CNN2D). Each algorithm was evaluated using accuracy, precision, recall, F-Score, confusion matrix, and AUC-ROC analysis. Results show that Logistic Regression and CNN2D achieved the highest performance, both exceeding 99% accuracy, followed by Random Forest (96%) and SVM (94%). These results confirm that machine learning and deep learning models are highly effective at distinguishing normal IoT traffic from cyber-attack traffic, enabling proactive detection, automated incident response, and an overall improvement in the security posture of IoT deployments.