Abstract
IoT devices with limited resources, and in the absence of gateways, become vulnerable to various attacks, such as denial of service (DoS) and man-in-the-middle (MITM) attacks. Intrusion detection systems (IDS) are designed to detect and respond to these threats in IoT environments. While machine learning-based IDS have typically been deployed at the edge (gateways) or in the cloud, in the absence of gateways, the IDS must be embedded within the sensor nodes themselves. Available datasets mainly contain features extracted from network traffic at the edge (e.g., Raspberry Pi/computer) or cloud servers. We developed a unique dataset, named as Intrusion Detection in the Smart Homes (IDSH) dataset, which is based on features retrievable from microcontroller-based IoT devices. In this work, a Tree-based IDS is embedded into a smart thermostat for real-time intrusion detection. The results demonstrated that the IDS achieved an accuracy of 98.71% for binary classification with an inference time of 276 microseconds, and an accuracy of 97.51% for multi-classification with an inference time of 273 microseconds. Real-time testing showed that the smart thermostat is capable of detecting DoS and MITM attacks without relying on a gateway or cloud.
Original language | English |
---|---|
Article number | 7320 |
Journal | Sensors |
Volume | 24 |
Issue number | 22 |
Early online date | 16 Nov 2024 |
DOIs | |
Publication status | Published - 16 Nov 2024 |
Keywords
- CatBoost
- DoS
- embedded ML
- intrusion detection system
- MITM
- TinyML
ASJC Scopus subject areas
- Analytical Chemistry
- Information Systems
- Atomic and Molecular Physics, and Optics
- Biochemistry
- Instrumentation
- Electrical and Electronic Engineering