Real-time attack analysis and defense technology for IoT

• Main Author: Sie, Jovan
• Other Authors: Liu Yang
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; IoT; Cybersecurity
• Online Access: https://hdl.handle.net/10356/181159
• Institution: Nanyang Technological University

The research addresses the increase in cyber attacks on IoT networks and explores the use of multi-class classification techniques to improve current iterations of intrusion detection systems. The research used an innovative dataset known as TON_IoT, to perform feature extraction in identifying network attacks, and for training and testing of 6 different classification models, which were then filtered to be implemented into an intrusion detection system for real-world testing. Features for each attack type in the dataset were analysed and rated in importance on distinguishing the individual attacks from benign traffic. The 6 classification models yielded varying results, with one attaining a value of 0.98 in accuracy. The models were tested against real-world data attaining an accuracy of 0.68. The study proposes the use of multi-class classification in performing anomaly-based intrusion detection systems, to create accurate and tailored response for different attack types.