MoT-Zb Profiling Dataset

This repository accompanies the data article A Comprehensive Dataset for Device Fingerprinting and Traffic Classification in Zigbee and Matter over Thread Networks.

It contains the raw captures, metadata, and ground truth used in the study. It also includes Python notebooks that reproduce the feature extraction and classification workflow described in the article.

Dataset Summary

• 6 continuous captures, 12 hours each
• 2 protocols: Matter over Thread and Zigbee
• 3 topology settings: A, B, and C
• Ground truth artifacts for each capture:
  • packet capture files (.pcapng)
  • capture metadata and device tables (CaptureX_description.md)
  • state-change logs (CaptureX_log.csv)
  • command logs (CaptureX_output.csv)
  • topology snapshots (CaptureX_graph.svg)

Topology Mapping

• Topology A: baseline full-mesh layout
• Topology B: distributed residential layout 1 (multi-hop)
• Topology C: distributed residential layout 2 (multi-hop)

Reproducibility Workflow

The notebooks are organized as follows:

1. Feature_Extraction.ipynb reads the packet captures and produces 5-second window feature tables.
2. Classification.ipynb loads the extracted features and runs the Random Forest classification example used in the paper.

Notes on Metadata

• CaptureX_description.md is the primary human-readable source for each capture.
• For Matter over Thread, the capture description includes the device table, capture start time, decryption key, and embedded topology snapshots.
• For Zigbee, topology snapshots are exported separately as JSON files using zha_toolkit.