RoFiRe: Robust Website Fingerprinting on Real-World Tor Traffic via Improved Augmentation and Normalization

[Paper Link (will be provided soon)]

Haeseung Jeon*, Sujin Kim*, Nate Mathews, Ho Sung Kang, and Se Eun Oh†

*Equally credited authors. †Corresponding author.

Note

This is the authors' RoFiRe attack model code proposed in RoFiRe: Robust Website Fingerprinting on Real-World Tor Traffic via Improved Augmentation and Normalization work, presented in the WWW'26.

1. Environment

We utilized a single NVIDIA RTX A6000 GPU (40GB VRAM) in a Ubuntu 20.04 server with 1.0 TB RAM, 7TB SATA SSDs, two NVMe SSDs, and CUDA 11.4.

2. Getting Started

2-1. Project Structure

This repository provides the full pipeline for feature extraction, augmentation, model training, and evaluation.

RoFiRe/
│
├── auto_aug.py              # Entry point: experiment settings, hyperparameters, paths
├── logger_config.py         # Logger setup and configuration utilities
│
├── features/
│   └── features.py          # Apply augmentation rule
│
└── model/
    ├── DF.py                # Model architecture (Deep Fingerprinting-based)
    ├── eval.py              # Draw PR curve for open-world evaluation
    └── main.py              # Training and evaluation pipeline

2-2. Install Dependencies

Installs all the necessary Python libraries and packages defined in the requirements.txt file.

pip install -r requirements.txt

2-3. Running RoFiRe

Starting point:

python auto_aug.py

Modify or add augmentation rules here: features/features.py

4. Contacts

Please contact us if you have any questions about RoFiRe.

• Haeseung Jeon, haeseungjeon@ewha.ac.kr
• Sujin Kim, sml09181@ewha.ac.kr
• Se Eun Oh, seoh@ewha.ac.kr

5. Reference

• DeepFingerprinting
• RMSNorm
• LabelSmoothing