Anomaly based network intrusion detection for IoT attacks using deep learning technique

Abstract

Internet of Things (IoT) applications are growing in popularity for being widely used in many real-world services. In an IoT ecosystem, many devices are connected with each other via internet, making IoT networks more vulnerable to various types of cyber attacks, thus a major concern in its deployment is network security and user privacy. To protect IoT networks against various attacks, an efficient and practical Intrusion Detection System (IDS) could be an effective solution. In this paper, a novel anomaly-based IDS system for IoT networks is proposed using Deep Learning technique. Particularly, a filter-based feature selection Deep Neural Network (DNN) model where highly correlated features are dropped has been presented. Further, the model is tuned with various parameters and hyper parameters. The UNSW-NB15 dataset comprising of four attack classes is utilized for this purpose. The proposed model achieved an accuracy of 84%. Generative Adversarial Networks (GANs) were used to generate synthetic data of minority attacks to resolve class imbalance issues in the dataset and achieved 91% accuracy with balanced class dataset.