A deep learning approach for intrusion detection in Internet of Things using bi-directional long short-term memory recurrent neural network

  • Bipraneel Roy

Western Sydney University thesis: Master's thesis

Abstract

Internet-of-Things connects every 'thing' with the Internet and allows these 'things' to communicate with each other. IoT comprises of innumerous interconnected devices of diverse complexities and trends. This fundamental nature of IoT structure intensifies the amount of attack targets which might affect the sustainable growth of IoT. Thus, security issues become a crucial factor to be addressed. A novel deep learning approach have been proposed in this thesis, for performing real-time detections of security threats in IoT systems using the Bi-directional Long Short-Term Memory Recurrent Neural Network (BLSTM RNN). The proposed approach have been implemented through Google TensorFlow implementation framework and Python programming language. To train and test the proposed approach, UNSW-NB15 dataset has been employed, which is the most up-to-date benchmark dataset with sequential samples and contemporary attack patterns. This thesis work employs binary classification of attack and normal patterns. The experimental result demonstrates the proficiency of the introduced model with respect to recall, precision, FAR and f-1 score. The model attains over 97% detection accuracy. The test result demonstrates that BLSTM RNN is profoundly effective for building highly efficient model for intrusion detection and offers a novel research methodology.
Date of Award2018
Original languageEnglish

Keywords

  • intrusion detection systems (computer security)
  • Internet of things
  • machine learning
  • neural networks (computer science)
  • computer networks
  • security measures

Cite this

'