Purdue-TC Controlled Dataset Collection

This repository provides scripts and configuration files to collect controlled network traffic datasets for traffic classification research.
The datasets capture application flows under systematically varied network conditions (capacity, delay, background traffic, user behavior), enabling reproducible experiments on data augmentation and traffic classification methods.

---

System Resources

Experiments were conducted on a high-performance server with the following specifications:

• CPU: 2 × Intel E5-2630 v3 (Haswell, 8 cores each, 2.40 GHz)
• RAM: 128 GB ECC DDR4 (1866 MHz)
• Storage:
  • 2 × 1.2 TB 10K RPM SAS HDDs
  • 1 × 480 GB Intel DC S3500 SSD
• OS: Ubuntu 20.04 (UBUNTU20-64-STD)

---

Dataset Description

The dataset consists of flows from five application types:

1. Video Streaming (VS): DASH streaming of Big Buck Bunny & Elephants Dream at multiple resolutions.
2. Video Conferencing (VC): Two-way audio-video calls with aiortc.
3. Web Browsing: Selenium-based scripted browsing of top-500 websites with realistic wait times.
4. FTP: File transfers generated from synthetic file distributions.
5. Email: POP3 client/server fetching messages with polling.

Each application is run across controlled conditions of:

• Bottleneck link capacities: 5, 10, 100, 200 Mbps
• Round-trip times (RTT): 1, 10, 50, 75 ms
• Background traffic: On/Off UDP traffic at different rates
• User behavior: Varying think times for browsing sessions

These systematic variations produce a rich dataset for evaluating traffic classification robustness.

---

Setup & Installation

1. Install Dependencies

Run the provided installation scripts in order:

sudo bash install_script_part1.sh
sudo bash install_script_part2.sh

These scripts install Docker, Containernet, traffic generators, and supporting libraries.

2. Configure Dataset Paths

Update the base directory in controlled_config.ini:

[PATHS]
base = /path/to/REPO

This path will be used for mounting volumes such as FTP, Web, Video Streaming, and Conferencing traces. This path is where the repository is located.

3. Run Experiments

Start the experiment pipeline:

sudo ./run_experiments.sh

This script will:

• Launch the dumbbell topology in Containernet
• Deploy client/server containers with applications
• Execute workloads across configurations
• Save packet traces into the configured dataset paths

---

Reference

If you use this dataset in your work, please cite:

---

License

This project is released for academic and research use. Please check the repository’s license file for details.