🔐 DotScramble

Advanced Image Privacy Studio

A powerful, modular image privacy tool with AI evasion, face detection, multiple effects, batch processing, and real-time preview — built on PySide6.

Quick Start • Features • AI Evasion • Usage • Installation • Contributing

🚀 Quick Start

No Python installation required! Download and run the standalone executable.

🐧 Linux

Install System Dependencies (for text detection features):
BASH
```
sudo apt install tesseract-ocr tesseract-ocr-eng
```

Download your preferred package from the Releases Page:

Package	Use case
`DotScramble-Linux-x86_64.AppImage`	Universal — run on any distro
`DotScramble-Linux-x86_64.deb`	Debian / Ubuntu / Kali — integrates with app menu
`DotScramble-Linux-x86_64.pkg.tar.zst`	Arch Linux / Manjaro — install with pacman

AppImage:

BASH

chmod +x DotScramble-Linux-x86_64.AppImage
./DotScramble-Linux-x86_64.AppImage

Or .deb:

BASH

sudo dpkg -i DotScramble-Linux-x86_64.deb
dotscramble

Or Arch (.pkg.tar.zst):

BASH

sudo pacman -U DotScramble-Linux-x86_64.pkg.tar.zst
dotscramble

Note: Text detection requires Tesseract OCR. Without it, you'll see a warning but all other features work normally.

📦 Download Latest Release

🔏 Security & Integrity Verification

All release binaries are signed with GPG. You can verify authenticity before running anything.

🔑 Step-by-step verification guide

1. Import the signing public key

BASH

# Option A — from the release assets (download public_key.asc first)
gpg --import public_key.asc

# Option B — from Ubuntu Keyserver
gpg --keyserver keyserver.ubuntu.com --recv-keys 0D9B71AF1791DA36

2. Verify the key fingerprint

CODE

Key ID    : 0D9B71AF1791DA36
Fingerprint: 7D06 4BC6 C9E2 34B8 948D D12D 0D9B 71AF 1791 DA36
UID       : FreeRave <kareem209907@gmail.com>

BASH

gpg --fingerprint 0D9B71AF1791DA36

3. Verify a release binary

BASH

# Verify AppImage
gpg --verify DotScramble-Linux-x86_64.AppImage.asc DotScramble-Linux-x86_64.AppImage

# Verify .deb
gpg --verify DotScramble-Linux-x86_64.deb.asc DotScramble-Linux-x86_64.deb

A Good signature message confirms the file is authentic and unmodified. ✅

4. Verify SHA-256 checksums

BASH

# Download SHA256SUMS from the release, then:
sha256sum -c SHA256SUMS

✨ Features

🛡️ Privacy & AI Evasion (New in v1.4.0)

🤖 AI Evasion System — SPSA black-box adversarial perturbations defeat AI face recognition models without any visible artifact at low strength settings
🎭 Layered Protection — Adversarial noise + visual blur + EXIF strip applied in sequence
⚙️ Evasion Strength Slider — Subtle / Balanced / Maximum presets (ε 0.03 → 0.12)
🔌 Pluggable Proxy Model — Swap in your own face recognition model as the optimization target

🎯 Detection Modes

🎭 Face Detection — Detect and obscure faces (MediaPipe + Haar cascade ensemble)
👁️ Eye Detection — Target specific eye regions
🧍 Full Body Detection — Detect entire person silhouettes
🚗 License Plate Detection — Auto-identify vehicle plates
📝 Text Detection (OCR) — Find and censor text with Tesseract
✏️ Manual Selection — Draw custom regions
🌍 Full Image — Apply effects to the entire image

🌍 Localization & UI

Native Arabic RTL interface with dynamic language switching
16 built-in themes: Cyberpunk, One Dark, Dracula, Nord, Catppuccin Mocha, Rosé Pine, and more
Responsive layout with Wayland-compatible window management

🎨 Privacy Effects

🌫️ Gaussian Blur — Smooth, professional blur
🔲 Pixelation — Classic pixel censoring
⬛ Black Bar — Solid rectangular censor
🎭 Gradient Fade — Artistic gradient transition
🔳 Mosaic — Decorative tile pattern
❄️ Frosted Glass — Translucent glass effect
🎨 Oil Paint — Artistic painting style

💎 Core Capabilities

MVC Architecture for maximum stability and extensibility
Auto-Save Database via SQLite — settings and state persist across sessions
Real-time Preview — see effects live as you adjust sliders
Batch Processing — process entire folders in one go
Undo/Redo History — full state management
EXIF Metadata Spoofing & Stripping — modify or remove GPS, camera model, timestamps
Secure Auto-Update — Ed25519-signed binary verification before any hot-swap

💎 Advanced Capabilities

Feature	Description
🤖 AI Evasion	SPSA adversarial perturbations — defeats embedding-based face recognition
⚡ Real-time Preview	See effects instantly as you adjust parameters
📦 Batch Processing	Process hundreds of images automatically
↶↷ Undo/Redo	Full history management with keyboard shortcuts
💾 Presets System	Save and load your favorite effect configurations
🔍 Comparison View	Side-by-side before/after comparison
⌨️ Keyboard Shortcuts	Lightning-fast workflow with hotkeys
🎚️ Opacity Control	Blend effects with original image
🔄 Smart Auto-Update	Cryptographically verified background updates

🤖 AI Evasion System

New in v1.4.0 — blurring a face is no longer enough.

Modern AI face recognition systems work on mathematical feature vectors, not visual legibility. A face blurred enough to fool a human eye can still produce a recognizable embedding inside a convolutional neural network.

DotScramble's AI Evasion System applies adversarial perturbations — mathematically-optimized noise patterns that corrupt these feature vectors — using SPSA (Simultaneous Perturbation Stochastic Approximation), a black-box optimization algorithm that requires no access to the target model's weights or gradients.

Original → Adversarial perturbation (SPSA) → Visual blur → EXIF strip → Protected image

Evasion Strength Presets

Preset	ε value	Visual impact	Evasion strength
Subtle	0.03	Nearly invisible	Moderate
Balanced (default)	0.05	Very subtle grain	Good
Maximum	0.12	Visible texture	Very strong

Benchmark Results (LFW dataset, 200 images)

Mode	AI Match Rate	Processing time
No effect	98.5%	—
Blur only	61.3%	12ms
Blur + AI Evasion (Balanced)	8.9%	~92s

⚠️ Note: Export as PNG to preserve adversarial perturbations. JPEG compression partially degrades them. The app warns you if you attempt JPEG export with evasion enabled.

📖 Usage

Basic Workflow

MERMAID

graph LR
    A[📁 Load Image] --> B[🎯 Select Mode]
    B --> C[🎨 Choose Effect]
    C --> D[🤖 Enable AI Evasion?]
    D --> E[⚙️ Adjust Settings]
    E --> F[✨ Apply Effect]
    F --> G[💾 Save as PNG]

🤖 AI Evasion Mode

Load your image and select Face Detection mode
Toggle "🤖 AI Evasion" in the processing panel
Choose your evasion preset (Subtle / Balanced / Maximum)
Click "✨ Apply Effect" — a progress bar shows optimization progress
Save as PNG (not JPEG) to preserve the adversarial perturbations

📷 Manual Selection Mode

Select "✏️ Manual Selection" from detection modes
Click and drag on the image to draw rectangles
Create multiple regions as needed
Click "✨ Apply Effect" to process all selected areas
Use "🗑️ Clear Selections" to reset and start over

⚡ Smart Settings

Auto-Save: Changing a slider automatically saves that value for next time.

Themes: Go to View → Themes to change the app look (16 themes available).

Language: Go to View → Language to switch between English and Arabic (full RTL support).

📦 Batch Processing

Click "📦 Batch Process" or press Ctrl+B
Select multiple images from your folders
Choose output directory for processed images
Configure detection mode and effect settings
Click Start and let it run automatically

💾 Presets Management

Configure your perfect settings (effect, strength, opacity, etc.)
Go to Menu → Presets → Save Current Settings
Name your preset (e.g., "Face Blur Strong", "Plate Pixelate")
Load anytime for consistent, repeatable results

⌨️ Keyboard Shortcuts

Shortcut	Action	Shortcut	Action
`Ctrl+O`	Open Image	`Ctrl+S`	Save Result
`Ctrl+Z`	Undo	`Ctrl+Y`	Redo
`Ctrl+P`	Apply Effect	`Ctrl+D`	Clear Selections
`Ctrl+B`	Batch Process	`Ctrl+Q`	Quit Application

🎚️ Effect Parameters

🌫️ Blur Strength

Range: 15–199
Odd numbers only for optimal results
Higher = stronger blur
Recommended: 31–51 for faces

🔲 Pixel Block Size

Range: 5–50 pixels
Higher = heavier censoring
Lower = subtle effect
Recommended: 15–25 for balance

🤖 Evasion Strength

Presets: Subtle / Balanced / Maximum
Higher = stronger AI evasion
Higher = longer processing time
Export as PNG for best results

💻 Installation

Option 1: Standalone Executable (Recommended)

No dependencies required! Download from the Releases Page.

Option 2: From Source

Click to expand installation steps

Prerequisites

Python 3.10 or higher
pip package manager

Steps

BASH

# Clone the repository
git clone https://github.com/kareem2099/DotScramble.git
cd DotScramble

# Create a virtual environment (recommended)
python3 -m venv venv
source venv/bin/activate  # Linux/macOS

# Install dependencies
pip install -r requirements.txt

# Run the application
python src/main.py

Core Dependencies

CODE

PySide6        — Qt6 GUI framework (LGPL v3)
opencv-python  — Computer vision
numpy          — Numerical computation
Pillow         — Image I/O
pytesseract    — OCR text detection
mediapipe      — Face landmark detection
requests       — Auto-update HTTP client
cryptography   — Ed25519 signature verification

📁 Project Structure

CODE

DotScramble/
│
├── 📋 requirements.txt           # Python dependencies
├── ⚙️  setup_license.py          # Cython compilation for license module
├── ⚙️  build.py                  # Secure staging build pipeline
│
├── 🧠 core/                      # System integrations & engines
│   ├── adversarial_engine.py     # SPSA black-box AI evasion engine  ← NEW
│   ├── auto_updater.py           # Ed25519-signed auto-updater
│   ├── image_picker.py           # Unicode/Arabic-safe file browser
│   ├── batch_processor.py        # Multi-threaded batch processor
│   ├── text_detector.py          # Tesseract OCR wrapper
│   └── metadata_spoofer.py       # EXIF metadata modifier & stripper
│
├── 🖼️  gui/                      # Additional GUI components
│   ├── metadata_dialog.py        # EXIF editor dialog
│   ├── metadata_presets.py       # EXIF device presets
│   └── metadata_report.py        # EXIF diagnostic report dialog
│
├── 📦 src/                       # Core MVC Application
│   ├── 📄 main.py                # Entry point
│   ├── ⚙️  config.py             # App config, theme definitions & URLs
│   ├── 🧠 controllers/           # MVC controllers
│   ├── 🧠 models/                # Processing engines & detectors
│   ├── 🧠 views/                 # RTL/LTR UI views & dialogs
│   └── 🧠 managers/              # Auth, DB, Theme, Locale managers
│
├── 🎨 assets/
│   ├── themes/themes.json        # 16 built-in color themes
│   └── icons/                    # Application icons
│
├── 💾 presets.json               # Saved effect presets (auto-generated)
└── 🗂️  backups/                  # Automatic image backups (auto-created)

🛠️ Customization

Adding Custom Effects

Click to see example code

Open core/image_processor.py
Add your effect method:

PYTHON

@staticmethod
def my_custom_effect(image, x, y, w, h):
    """
    Apply custom effect to image region.

    Args:
        image: Source image (numpy array, BGR)
        x, y:  Top-left corner coordinates
        w, h:  Width and height of region

    Returns:
        Processed region (numpy array)
    """
    region = image[y:y+h, x:x+w]
    processed = 255 - region  # Example: invert colors
    return processed

PYTHON

EFFECTS = {
    # ... existing effects ...
    'my_custom_effect': '🌟 My Custom Effect'
}

Add to the GUI effect selection dropdown.

Plugging In a Custom AI Evasion Model

Click to see example code

PYTHON

from core.adversarial_engine import SPSAAdversarialEngine
import numpy as np

# Implement the EvasionProxy protocol with your model
def my_loss_fn(perturbed_img: np.ndarray) -> float:
    """Higher return value = more adversarial."""
    embedding = my_model.get_embedding(perturbed_img)
    if embedding is None:
        return 1.0
    return float(1.0 - np.dot(original_embedding, embedding))

engine = SPSAAdversarialEngine(epsilon=0.05, num_iters=150)
result = engine.generate(face_crop, loss_fn=my_loss_fn)

🤝 Contributing

We welcome contributions! Here are some ideas to get started:

🎯 Ideas for Contributors

🎥 Video processing support
⚡ GPU acceleration (CUDA / Metal) for AI evasion
🌐 Web-based interface
📱 Mobile app version
🧠 Additional detection models (YOLO, MediaPipe Pose)
🔄 Batch undo/redo
📊 Processing statistics dashboard
🌍 Additional language localizations

📝 How to Contribute

Fork the repository
Create your feature branch
BASH
```
git checkout -b feature/AmazingFeature
```

Commit your changes

BASH

git commit -m 'feat: add AmazingFeature'

Push to the branch
BASH
```
git push origin feature/AmazingFeature
```
Open a Pull Request

📄 License

This project is licensed under the Apache License 2.0 — see the LICENSE file for details.

CODE

Apache License 2.0 — Free for personal and commercial use
PySide6 (Qt bindings) — LGPL v3 — compatible with Apache 2.0

🙏 Acknowledgments

OpenCV — Computer vision library
MediaPipe — Face landmark detection
PySide6 / Qt — Cross-platform GUI framework
Tesseract OCR — Open-source text recognition
Goodfellow et al. (2014) — Adversarial examples research that inspired the AI evasion system

📞 Support & Contact

Need Help?

Troubleshooting

Common Issues

Q: I see a Tesseract warning on startup? A: Text detection requires Tesseract OCR. Install it with:

BASH

sudo apt install tesseract-ocr tesseract-ocr-eng

Restart the application after installation.

Q: AI Evasion is taking very long? A: Adversarial optimization is CPU-intensive. Use the Subtle preset for faster processing (~45s per face vs ~90s for Balanced). GPU acceleration is planned for a future release.

Q: Face detection not working? A: Ensure good lighting and front-facing angles. Try Manual Selection as a fallback for difficult angles.

Q: Application won't start? A: Check that all dependencies are installed. Try running from source:

BASH

python src/main.py

Q: Batch processing is slow? A: Processing time depends on image size and effect complexity. Use Pixelation for faster processing, or disable AI Evasion for batch jobs.

Q: Check for Updates shows nothing? A: Ensure you're running the official release build (not from source). The update check requires an internet connection and will show "You are using the latest version" if already up to date.

⭐ Star this repository if you find it helpful!

Made with ❤️ by FreeRave for privacy protection

⬆ Back to Top

DotScramble

🔐 DotScramble

Advanced Image Privacy Studio

🚀 Quick Start

🐧 Linux

🔏 Security & Integrity Verification

1. Import the signing public key

2. Verify the key fingerprint

3. Verify a release binary

4. Verify SHA-256 checksums

✨ Features

🛡️ Privacy & AI Evasion (New in v1.4.0)

🎯 Detection Modes

🌍 Localization & UI

🎨 Privacy Effects

💎 Core Capabilities

💎 Advanced Capabilities

🤖 AI Evasion System

Evasion Strength Presets

Benchmark Results (LFW dataset, 200 images)

📖 Usage

Basic Workflow

⌨️ Keyboard Shortcuts

🎚️ Effect Parameters

🌫️ Blur Strength

🔲 Pixel Block Size

🤖 Evasion Strength

💻 Installation

Option 1: Standalone Executable (Recommended)

Option 2: From Source

Prerequisites

Steps

Core Dependencies

📁 Project Structure

🛠️ Customization

Adding Custom Effects

Plugging In a Custom AI Evasion Model

🤝 Contributing

🎯 Ideas for Contributors

📝 How to Contribute

📄 License

🙏 Acknowledgments

📞 Support & Contact

Troubleshooting

⭐ Star this repository if you find it helpful!

منتجات مشابهة