IoT Device Attack Detection using Deep Learning in Robust Framework for Securing Smart Systems

Automation, connectivity, and intelligence over the IoT has become mainstream in its contemporary metaphor. The exponential growth of IoT networks, however, not only has easily matured more vulnerabilities, especially in network layer, that is targets most of the time by sophisticated cyber-attacks as distributed denial of service (DDoS), man in the middle (MITM), and malware propagation. Whereas traditional security mechanisms are faced with both these challenges, they rarely succeed in addressing them in real time especially in a resource constrained IoT environment. Thus, in order to fill this gap, this study introduces a deep learning based intrusion detection framework able to classify different types of multi class network attacks in smart systems with high efficiency and low accuracy. With 20 numerical behavioral features, the model classifies the traffic into three classes: benign, moderate-level threat class, and high severity attack class. Through confusion matrix, ROC curves and precision recall curves, the performance of the model is evaluated and was able to get over 94% validation accuracy with balanced per class performance and minimal over fitting. Moreover, it is a practical solution to secure IoT ecosystems because it is designed to support real time implementation in low power devices. These researches provided contribution in order to further intelligent, lightweight, and scalable intrusion detection methods into smart environments for future proofing.