Discipline of Information Technology, College of Science & Engineering, James Cook University, Townsville, QLD 4811, Australia.
Sensors (Basel). 2022 Jul 27;22(15):5626. doi: 10.3390/s22155626.
The Smart Home (SH) has become an appealing target of cyberattacks. Due to the limitation of hardware resources and the various operating systems (OS) of current SH devices, existing security features cannot protect such an environment. Generally, the traffic patterns of an SH IoT device under attack often changes in the Home Area Network (HAN). Therefore, a Network-Based Intrusion Detection System (NIDS) logically becomes the forefront security solution for the SH. In this paper, we propose a novel method to assist classification machine learning algorithms generate an anomaly-based NIDS detection model, hence, detecting the abnormal SH IoT device network behaviour. Three network-based attacks were used to evaluate our NIDS solution in a simulated SH test-bed environment. The detection model generated by traditional and ensemble classification Mechanical Learning (ML) methods shows outstanding overall performance. The accuracy of all detection models is over 98.8%.
智能家居（SH）已成为网络攻击的诱人目标。由于当前 SH 设备的硬件资源和各种操作系统（OS）的限制，现有安全功能无法保护此类环境。通常，受到攻击的 SH IoT 设备的流量模式在家庭区域网络（HAN）中经常发生变化。因此，基于网络的入侵检测系统（NIDS）在逻辑上成为 SH 的前沿安全解决方案。在本文中，我们提出了一种新方法，以协助分类机器学习算法生成基于异常的 NIDS 检测模型，从而检测异常的 SH IoT 设备网络行为。在模拟的 SH 测试床环境中，使用三种基于网络的攻击来评估我们的 NIDS 解决方案。传统和集成分类机械学习（ML）方法生成的检测模型表现出出色的整体性能。所有检测模型的准确性均超过 98.8％。
