Skip to content

hyboxteam/Parda-Sheild

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 

Repository files navigation

🛡️ Parda-Shield

Parda-Shield is a lightweight, browser-based Progressive Web App (PWA) designed to protect human dignity in the age of generative AI. It "immunizes" your images before they ever touch the cloud, making them hostile to AI scrapers and manipulation engines.

"In an era where pixels are weapons, Parda is your digital armor." — HyBox Ideation Engine


🚀 Why Parda-Shield?

As AI models (Stable Diffusion, Deepfake generators, etc.) become more prevalent, the risk of unauthorized image manipulation—especially against vulnerable individuals—has skyrocketed. Parda-Shield addresses this at the First Principles level:

  • Proactive Defense: Don't just detect deepfakes; prevent them from being created accurately.
  • Zero-Budget Accessibility: No expensive servers, no subscriptions. It runs entirely in your browser.
  • Privacy by Design: Your photos never leave your device. All processing happens locally.

🛠️ How It Works (The "Shield" Mechanism)

Parda-Shield applies a dual-layered defense to every image:

1. Layer 1: Metadata Poisoning (The Decoy)

Most AI scrapers rely on EXIF data and metadata to categorize images. Parda-Shield:

  • Strips all original sensitive data (GPS, camera info, etc.).
  • Injects a Poison Payload: Categorizes the image as "Random Landscape" or "Abstract Pattern" to confuse automated datasets.

2. Layer 2: Adversarial Noise (The Disruptor)

Using the Canvas API, the app injects a subtle, high-frequency Gaussian Noise layer (1-2% intensity).

  • To Humans: The image looks identical.
  • To AI: The noise disrupts the "Latent Space" mapping, causing generative models to produce glitched or distorted results if they try to manipulate the face or features.

📦 Features

  • PWA Ready: Install it on your Android/iOS device directly from the browser.
  • Offline First: Works without an internet connection once loaded.
  • GitHub Pages Compatible: Zero hosting costs.
  • Privacy Guaranteed: 100% Client-side processing.

🛠️ Tech Stack

  • Core: HTML5, Tailwind CSS, Vanilla JavaScript.
  • Libraries: piexifjs (for metadata manipulation).
  • Deployment: GitHub Pages.

🎓 For Students & Developers

Parda-Shield is a part of the HyBox mission to empower zero-budget creators. We encourage you to:

  1. Fork this repo.
  2. Experiment with noise intensities in script.js.
  3. Contribute by adding new adversarial patterns (e.g., FGSM attacks).

⚖️ Ethics & Safety

Parda-Shield is built for protection, not deception. We are committed to human-AI interaction integrity. This tool is a defensive response to the misuse of uncensored AI models.


🤝 Contribution

Created with 💡 by Awais (Founder) & Aarav (Chief Ideator) at HyBox. Special thanks to the open-source community for making privacy accessible.

Let's build a safer digital world, one pixel at a time.

About

Parda-Shield: A privacy-first, zero-budget PWA to immunize images against AI manipulation, deepfakes, and unauthorized scraping using adversarial noise and metadata poisoning. Built for the HyBox mission.

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages