Identify devices and events from non-IP heterogeneous IoT network traffic

Area of Science:

• Computer Science
• Network Security

Background:

• Current methods for identifying Internet of Things (IoT) devices primarily focus on IP-based traffic, limiting their effectiveness with heterogeneous non-IP networks.
• Existing techniques struggle with diverse IoT protocols (e.g., ZigBee, Z-Wave) and insufficient traffic samples, hindering accurate device and event identification.

Purpose of the Study:

• To develop a novel approach for identifying IoT devices and events within non-IP heterogeneous IoT network traffic.
• To overcome limitations of existing methods by accommodating diverse IoT protocols and addressing the challenge of scarce traffic samples.

Main Methods:

• Proposed a novel approach leveraging IoT communication characteristics and module extensibility for identifying non-IP heterogeneous IoT traffic.
• Developed a heterogeneous sample extraction platform with an extensible structure to acquire raw sequence samples from ZigBee and Z-Wave traffic.
• Devised a sample identification framework to generate synthetic samples from raw data, enabling concurrent processing with raw samples via a dual-sequence network model.

Main Results:

• The proposed method significantly improves the identification of non-IP heterogeneous IoT traffic compared to baseline and state-of-the-art techniques.
• Achieved an average accuracy improvement of 29.7% over baseline models using only raw samples.
• Demonstrated superior performance with improvements of 22.1% in macro precision, 21.5% in macro recall, and 21.8% in macro F1-score over the latest method.

Conclusions:

• The novel approach effectively identifies IoT devices and events in non-IP heterogeneous networks, outperforming existing methods.
• The method's ability to generate synthetic samples and process diverse traffic enhances its robustness and accuracy in challenging IoT environments.
• The extensible platform design allows for future integration of additional IoT protocols, ensuring long-term applicability.