How do I run Fawkes on a GPU for faster processing?

Clone the repository, install required packages from setup.py, replace tensorflow with tensorflow-gpu, and run python3 fawkes/protection.py with arguments like --gpu and --batch-size. The binary package does not support GPU out of the box.

Is Fawkes better than blurring faces for privacy protection?

Fawkes uses subtle, invisible perturbations that preserve image quality while disrupting AI, unlike blurring which visibly degrades photos. However, blurring is simpler and more accessible, whereas Fawkes requires technical setup and may not be as universally effective.

Can Fawkes protect videos or only images?

Fawkes is designed for static images and does not support video processing natively. To protect videos, you'd need to extract frames, process them individually with Fawkes, and reassemble, which is not built-in and adds manual steps.

What's the difference between low, mid, and high modes in Fawkes?

Low mode adds minimal perturbation for basic protection with little visual change, mid balances strength and visibility, and high provides maximum protection with more noticeable alterations. Choose based on your privacy needs and image quality tolerance.

How long does it take to process 100 images with Fawkes on CPU?

At about 60 seconds per image on CPU, processing 100 images would take approximately 100 minutes or 1.7 hours. Using GPU with batch processing can cut this time significantly, but requires manual setup.

Does Fawkes work with all facial recognition systems?

Fawkes targets deep learning-based models common in research, but effectiveness may vary across systems. It's not guaranteed to protect against all facial recognition algorithms, especially as new models evolve.

Can I integrate Fawkes into my web application?

Not directly; Fawkes lacks an API or SDK. You would need to wrap the command-line tool in a server-side script, which adds complexity and may not scale well for real-time or user-facing applications.

Open-Awesome

fawkes

BSD-3-ClausePythonv0.3

A privacy protection tool that adds imperceptible pixel-level changes to photos to prevent unauthorized facial recognition.

Visit Website GitHub

5.5k stars501 forks0 contributors

What is fawkes?

Fawkes is a privacy protection tool that adds subtle, imperceptible pixel-level changes to personal photos to prevent unauthorized facial recognition systems from learning and identifying individuals. It addresses the growing concern of personal photos being scraped and used to train facial recognition models without consent. The tool provides different protection modes to balance privacy strength with visual fidelity.

Target Audience

Individuals concerned about online privacy, researchers studying privacy-preserving technologies, and organizations needing to protect biometric data from unauthorized AI training.

Value Proposition

Fawkes offers a practical, research-backed solution for proactive privacy protection against facial recognition systems, with multiple protection levels and batch processing capabilities. Unlike simple image obfuscation, it uses targeted perturbations that preserve image quality while disrupting AI model training.

Overview

Fawkes, privacy preserving tool against facial recognition systems. More info at https://sandlab.cs.uchicago.edu/fawkes

Use Cases

Best For

Protecting personal photos before sharing them on social media
Academic research on privacy-preserving technologies
Testing vulnerabilities in facial recognition systems
Preventing unauthorized use of biometric data in AI training
Adding privacy protection to image datasets
Understanding adversarial attacks on computer vision models

Not Ideal For

Applications requiring real-time image processing, such as live video streams or instant photo sharing.
Teams needing seamless integration into existing software pipelines without command-line dependencies.
Users who require verified and quantifiable protection levels, as effectiveness testing tools are incomplete.
Non-technical individuals looking for a simple, GUI-based tool for occasional use.

Pros & Cons

Pros

Research-Based Foundation

Built on peer-reviewed academic work published at USENIX Security 2020, ensuring credibility and a solid methodological backbone for privacy protection.

Adjustable Privacy Modes

Offers low, mid, and high settings to customize the trade-off between protection strength and image quality, as specified in the command-line options for flexible use.

Efficient Batch Processing

Supports processing multiple images simultaneously with configurable batch-size, optimizing performance on GPU machines for handling large datasets efficiently.

Proactive Privacy Strategy

Enables users to preemptively alter photos before online sharing, disrupting unauthorized facial recognition training while preserving visual fidelity for human viewers.

Cons

Complex GPU Setup

The standard binary package lacks GPU support; users must clone the repo, manually install tensorflow-gpu, and run from source, increasing setup complexity and barriers to performance.

Slow CPU Performance

Perturbation generation takes about 60 seconds per image on CPU, making it inefficient for large collections without GPU access, as noted in the tips section.

Unclear Effectiveness Metrics

The README admits that scripts to test protection effectiveness are not yet ready, leaving users uncertain about real-world security guarantees and reliability.

Limited Accessibility

Primarily a command-line tool with no graphical user interface or API, hindering adoption by non-developers and integration into automated or user-friendly systems.

Frequently Asked Questions

Related Projects

detectron2

Detectron2 is a platform for object detection, segmentation and other visual recognition tasks.

Stars34,348

Forks7,924

Last commit17 days ago

EasyOCR

Ready-to-use OCR with 80+ supported languages and all popular writing scripts including Latin, Chinese, Arabic, Devanagari, Cyrillic and etc.

Stars29,339

Forks3,558

Last commit4 months ago

imgaug

Image augmentation for machine learning experiments.

Stars14,736

Forks2,458

Last commit1 year ago

meshroom

Node-based Visual Programming Toolbox

Stars12,680

Forks1,206

Last commit2 days ago

Community-curated · Updated weekly · 100% open source

Found a gem we're missing?

Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.

Submit a project Star on GitHub