Ethernet Frame Physical-Layer Signal Dataset – 10BaseT

This dataset was developed to support research on network traffic classification using raw electrical signals captured at the physical layer. The core hypothesis behind this work is that different types of network traffic exhibit distinguishable patterns in their physical-layer waveforms due to variations in frame structure.

The dataset contains Ethernet frame signals corresponding to six widely used protocol types: DHCP, DNS, HTTP, ICMP, RTSP, and TLS. All data were collected in accordance with the 10Base-T Ethernet standard and are intended for research on signal-level network traffic classification.

To construct the raw signal dataset, protocol-specific packet capture (PCAP) files were collected from various sources. These files were then retransmitted over a 10Base-T Ethernet link, and the corresponding electrical signals were captured from the Ethernet cable using an oscilloscope. In total, 7421 unique signal samples (representing individual Ethernet frames) were extracted.

In addition to the raw signal files (provided in .csv format), the dataset also includes the original PCAP files used during acquisition.

Image Datasets (Visualization-Based):
In addition to the raw signal-level dataset, four separate image datasets were generated using different visualization techniques:
*vertical
*horizontal_zigzag
*spectrogram
*scalogram
Each visualization dataset contains 7,421 images, corresponding one-to-one with the raw signal files. These images were generated to support deep-learning-based classification experiments and are publicly shared together with the signal dataset.

MATLAB Script Packages:
Two MATLAB script packages are also provided to ensure full reproducibility:

1) Signal_Dataset_Generation_MATLAB_Scripts
Contains the scripts used for PCAP preprocessing, deduplication, packet-to-signal matching, automatic signal segmentation, and signal file labeling.

2) Image_Dataset_Generation_MATLAB_Scripts
Contains the scripts used to convert the raw 1-D signals into images using the four visualization techniques listed above.

Both script folders include a detailed script_information.txt file describing the purpose and functionality of each script.

Intended Use:
This dataset can be used for signal-based network traffic classification, physical-layer analysis, and deep learning research. It offers a novel perspective on traffic analysis beyond conventional packet- or flow-level features.