Toward a deep learning-driven intrusion detection approach for Internet of Things

Abstract Internet of Things (IoT) as a paradigm comes with a range of benefits to humanity. Domains of research for the IoT range from healthcare automation to energy and transport. However, due to their limited resources, IoT devices are vulnerable to various types of cyber attacks as carried out by the adversary. In this paper, we propose a novel intrusion detection approach for the IoT, through the adoption of a customised deep learning technique. We utilise a cutting-edge IoT dataset comprising IoT traces and realistic attack traffic, including denial of service, distributed denial of service, data gathering and data theft attacks. A feed-forward neural networks model with embedding layers (to encode high-dimensional categorical features) for multi-class classification, is developed. The concept of transfer learning is subsequently applied to encode high-dimensional categorical features to build a binary classifier based on a second feed-forward neural networks model. We obtain results through the evaluation of the proposed approach which demonstrate a high classification accuracy for both classifiers, namely, binary and multi-class.