Improving Network-Based Anomaly Detection in Smart Home Environment

Area of Science:

• Cybersecurity
• Internet of Things (IoT) Security
• Machine Learning Applications

Background:

• Smart Home (SH) environments are increasingly vulnerable to cyberattacks.
• Existing security solutions struggle with the diverse hardware and operating systems (OS) of SH IoT devices.
• Network traffic pattern anomalies in the Home Area Network (HAN) often indicate an attack on SH IoT devices.

Purpose of the Study:

• To propose a novel method for developing an anomaly-based Network-Based Intrusion Detection System (NIDS) for Smart Home environments.
• To leverage classification machine learning algorithms to generate an effective NIDS detection model.
• To detect abnormal network behavior indicative of compromised SH IoT devices.

Main Methods:

• Developed a novel method to assist classification machine learning algorithms in generating an anomaly-based NIDS detection model.
• Utilized traditional and ensemble classification Machine Learning (ML) methods.
• Evaluated the NIDS solution using three distinct network-based attacks in a simulated SH test-bed environment.

Main Results:

• The proposed method successfully assisted ML algorithms in generating NIDS detection models.
• All generated detection models demonstrated outstanding overall performance.
• The accuracy of all evaluated detection models exceeded 98.8% in identifying network anomalies.

Conclusions:

• The developed anomaly-based NIDS approach is highly effective for securing Smart Home IoT devices.
• Machine learning, particularly ensemble methods, shows significant promise for enhancing SH network security.
• The proposed solution offers a robust and accurate method for detecting cyber threats in Smart Home networks.