Fawkes Web Interface

The Fawkes Web Interface is a user-friendly application built with Flask, designed to leverage the powerful capabilities of Fawkes, a pioneering privacy-preserving technology that protects images from unauthorized facial recognition systems.

Overview

Fawkes enhances privacy by adding imperceptible modifications to images, or "cloaks," which render faces unrecognizable to facial recognition algorithms while keeping the visual integrity intact for human observers. The web interface simplifies this process, allowing users to effortlessly upload images, select cloaking modes, and apply privacy-preserving modifications through an intuitive drag-and-drop interface.

Key Features

• Easy-to-Use Interface: Upload images with ease and apply cloaking modifications with just a few clicks.
• Cloaking Modes: Choose from three different levels of cloaking:
  • Low Mode: Fastest processing with minimal protection (suitable for CPU/GPU use).
  • Mid Mode: Balanced protection and speed (GPU required).
  • High Mode: Maximum protection, ideal for sensitive images (GPU required).
• Multiple Image Format Support: The application supports PNG, JPEG/JPG, GIF, and BMP formats, with a maximum file size of 16MB.

How Fawkes Works

Fawkes employs advanced pixel-level perturbations that are invisible to the human eye but disrupt facial recognition systems, ensuring privacy while allowing for general usage. The modifications are designed to remain effective even after image compression and social media uploads, providing a robust defense against feature extraction by facial recognition tools.

Usage Instructions

Upon launching the Flask application, users can:

1. Upload an image through drag-and-drop or file selection.
2. Select the desired cloaking mode based on speed and protection needs.
3. Initiate the cloaking process and monitor the status.
4. Download the cloaked image once processing is complete.

Performance Optimization

To achieve optimal performance, it is recommended to:

• Utilize a GPU for faster image processing.
• Resize larger images for quicker processing times.
• Start with Low mode for testing before moving to High mode for final images.

Technical Specifications

The application is best run with Python 3.8.20 due to dependencies on specific versions of TensorFlow and complies with the requirements for GPU support, including having appropriate CUDA drivers installed.

Acknowledgments

• Fawkes Development: Created by Shawn Shan and the SANDLab at the University of Chicago.
• Original Research: For in-depth technical understanding, refer to the original Fawkes paper.

Important Note

This web interface serves as a tool for privacy protection and is intended for educational purposes. Users are responsible for ensuring compliance with all applicable laws and regulations regarding privacy protection.