🔐 AuthMark: Cryptographically Verifiable Image Watermarking

Securing digital authenticity one pixel at a time 🖼️✨

Ever wondered if that stunning image online is authentic or AI-generated? AuthMark solves this by embedding cryptographically verifiable watermarks directly into images using advanced DCT (Discrete Cosine Transform) techniques. It's like a digital fingerprint that's nearly impossible to forge! 🕵️‍♂️

🚀 What Makes AuthMark Special?

🔒 Cryptographic Security

• Ed25519 Digital Signatures: Each watermark is cryptographically signed and verifiable
• Tamper Detection: Know instantly if an image has been modified
• Non-repudiation: Prove the authenticity of your images beyond doubt

⚡ Lightning Fast Performance

• Multi-core Processing: Leverages all your CPU cores for blazing speed
• Optimized AI Detection: ~0.2s average processing time with smart fallbacks
• Smart Memory Management: Handles large images without breaking a sweat

🛡️ Robust Against Attacks

• JPEG Compression: Survives quality reduction and re-encoding
• Geometric Transforms: Resistant to cropping, rotation, and scaling
• Noise Attacks: Maintains integrity through various image manipulations

🤖 Advanced AI Detection System

• Hierarchical Detection: DMimage → Lightweight → Statistical fallback
• Smart Model Management: Auto-detects model reliability and falls back gracefully
• Specialised Models: DMimageDetection for diffusion models, lightweight for general detection
• Real-time Processing: ~0.2s per image with intelligent caching

🎯 Quick Start

Installation

# Clone the magic ✨
git clone https://github.com/pottsie283/authmark.git
cd authmark/watermarking

# Set up your environment 🏠
python -m venv AuthVenv
source AuthVenv/bin/activate  # On Windows: AuthVenv\Scripts\activate

# Install dependencies 📦
pip install -r requirements.txt

Launch the Application

# Start the Flask server 🚀
flask --app app.py run

# Visit the web interface 🌐
# http://localhost:5000/

💡 Usage Examples

🖥️ Web Interface (Easiest Way)

1. Upload & Watermark: Drop your image, enter your private key, and get a cryptographically signed watermarked image
2. Verify Authenticity: Upload any image to check if it contains a valid watermark
3. Robustness Testing: See how your watermarks survive various attacks
4. A/B Study: Test your ability to distinguish watermarked vs. original images

🔌 API Usage (For Developers)

Performance Monitoring:

# Check system performance 📊
curl http://localhost:5000/api/performance/stats

# Thread pool utilisation 🧵
curl http://localhost:5000/api/threadpool/stats

🔑 Key Generation

from core.crypto_utils import generate_key_pair

# Generate your cryptographic keys 🗝️
private_key_pem, public_key_pem = generate_key_pair()
print("Your keys are ready! 🎉")

🏗️ Architecture Overview

AuthMark/
├── 🌐 Web Interface (Flask + HTML/CSS/JS)
├── 🔌 REST API (Flask-RESTX with Swagger docs)
├── 🔐 Core Engine
│   ├── DCT Watermarking (Frequency domain embedding)
│   ├── Cryptographic Binding (RSA signatures)
│   ├── AI Anomaly Detection (Multi-method analysis)
│   └── Robustness Testing (Attack simulation)
├── ⚡ Performance Layer
│   ├── Multi-core Processing
│   ├── Thread Pool Management
│   └── Real-time Monitoring
└── 🧪 Testing & Evaluation Suite

🎮 Interactive Features

🔬 Robustness Laboratory

Test your watermarks against real-world scenarios:

• JPEG Compression (different quality levels)
• Geometric Attacks (rotation, scaling, cropping)
• Noise Injection (Gaussian, salt & pepper)

🧠 A/B Perceptual Study

Challenge yourself:

• Can you spot the watermarked image?
• 10 randomised questions
• Detailed performance metrics
• Perfect for research and validation

📊 Performance Dashboard

Monitor everything in real-time:

• Processing speeds and bottlenecks
• Memory usage optimisation
• Thread pool utilisation
• Success/failure rates

🎯 Key Features

✅ Invisible Watermarks: Imperceptible to human eyes (high PSNR/SSIM)
✅ Cryptographic Verification: RSA-2048 digital signatures
✅ Multi-format Support: JPEG, PNG, and more
✅ Batch Processing: Handle multiple images efficiently
✅ Attack Simulation: Built-in robustness testing
✅ Advanced AI Detection: Hierarchical detection with smart fallbacks
✅ Performance Monitoring: Real-time metrics and optimization
✅ RESTful API: Easy integration with other systems
✅ Interactive Web UI: No command-line expertise required

⚠️ Important Limitations

🚫 Video Support: Currently image-only (video watermarking is future work)
🚫 Key Management: In-memory only for demo purposes
🚫 GPS Data: Sensitive EXIF fields are excluded for privacy
🚫 Extreme Attacks: Some aggressive modifications may break watermarks
🚫 Real-time Video: Not optimised for live video streams

📋 What's Not Included

• Production Key Storage: Use proper HSMs or key management services
• User Authentication: Basic demo authentication only
• Cloud Deployment: Configured for local development
• Commercial Licensing: Research/educational use focused

🔬 Research Applications

Perfect for:

• Digital Forensics: Prove image authenticity in investigations
• Copyright Protection: Protect your creative works
• Academic Research: Study watermarking robustness and perceptual quality
• AI Detection Studies: Research synthetic media identification
• Performance Analysis: Benchmark cryptographic watermarking systems

🏆 Performance Benchmarks

Operation	Time	Details
AI Detection (DMimage)	~0.1s	Specialized diffusion model detection
AI Detection (Lightweight)	~0.2s	General AI content detection
AI Detection (Statistical)	~0.05s	Fallback analysis when models unavailable
Watermark Embedding	~2-5s	Multi-core DCT processing
Watermark Verification	~1-3s	Optimized extraction

🎯 Detection Accuracy

• Synthetic Images: 100% detection rate (DMimage model)
• Authentic Images: 60% correct classification (lightweight model)
• Smart Fallback: Automatic model reliability detection

🤝 Contributing

I'd love your help! Whether you're:

• 🐛 Finding bugs
• 💡 Suggesting features
• 📖 Improving documentation
• 🧪 Adding test cases
• ⚡ Optimising performance

Check out the issues page and join the conversation!

📜 License

MIT License - see LICENSE for details. Feel free to use this in your research, but please cite our work! 🙏

🎓 Academic Context

This system was developed as part of advanced cryptographic research into verifiable digital media authenticity. It demonstrates practical applications of:

• Frequency-domain watermarking techniques
• Cryptographic binding and non-repudiation
• Multi-core optimisation strategies
• Hierarchical AI detection systems
• Perceptual quality assessment

🌟 Star This Repo

If AuthMark helped you secure digital authenticity or advance your research, please give it a star! ⭐ It helps others discover this tool and motivates continued development.

---

Built with ❤️ for digital authenticity and security

"In a world of deepfakes and digital deception, AuthMark stands as a beacon of verifiable truth." 🔍✨