Toward a Deep Learning-Driven Intrusion Detection Approach for Internet of Things
Internet of Things (IoT) has brought along immense benefits to our daily lives encompassing a diverse range of application domains that we regularly interact with, ranging from healthcare automation to transport and smart environments. However, due to the limitation of constrained resources and computational capabilities, IoT networks are prone to various cyber attacks. Thus, defending the IoT network against adversarial attacks is of vital importance. In this paper, we present a novel intrusion detection approach for IoT networks through the application of a deep learning technique. We adopt a cutting-edge IoT dataset comprising IoT traces and realistic attack traffic, including denial of service, distributed denial of service, reconnaissance and information theft attacks. We utilise the header field information in individual packets as generic features to capture general network behaviours, and develop a feed-forward neural networks model with embedding layers (to encode high-dimensional categorical features) for multi-class classification. The concept of transfer learning is subsequently adopted to encode high-dimensional categorical features to build a binary classifier. Results obtained through the evaluation of the proposed approach demonstrate a high classification accuracy for both binary and multi-class classifiers.