AEE Protocol v0.2.3 - Scientific Watermarking

🎉 AEE Protocol v0.2.3 - Scientific Watermarking Release

After rigorous statistical validation (50,000+ trials), we present v0.2.3 with scientific threshold calculation and configurable false positive rates.

🔬 Key Innovations

✅ Scientific Threshold Calculation

• No more heuristic guessing
• Mathematically guaranteed false positive rates
• Threshold = z_score / √(n-3) where z_score comes from your desired FPR

✅ Configurable Security

• Choose your FPR: 0.1%, 1%, 2%, 5%
• Predictable, reliable performance
• Transparent operation

✅ Empirical Validation (50k+ trials)

• FPR 2% target → 1.88% observed
• FPR 0.1% target → 0.04% observed
• 44% survival with 20% Gaussian noise
• 100% detection with no noise

📊 Performance Summary

Configuration	Target FPR	Observed FPR	Noise Resistance (20% σ)
Balanced (strength=0.5)	2.00%	1.88%	44% survival
Strict (strength=0.5)	1.00%	0.93%	42% survival
Very Strict (strength=0.5)	0.10%	0.04%	8% survival

🚀 Quick Start

from aeeprotocol.sdk.client import AEEClient

# For general monitoring (recommended)
client = AEEClient(
    user_id=123456,
    strength=0.5,      # Balanced distortion/robustness
    target_fpr=0.02    # 2% false positive rate
)

# Watermark your embeddings
marked, proof = client.watermark(your_embedding)

# Verify later (blind detection)
result = client.verify(suspicious_embedding)
📈 What's Changed
v8.3 Engine: Scientific threshold calculation

Configurable FPR: Choose your security level

Improved SDK: Better API and error handling

Full Validation: 50,000+ test cases

Professional Documentation: Complete with results

📚 Documentation
[README.md](https://github.com/ProtocoloAEE/aee-protocol#readme) - Complete guide

[VALIDATION.md](https://github.com/ProtocoloAEE/aee-protocol/blob/main/VALIDATION.md) - Test methodology

🔧 Installation
bash
pip install aeeprotocol
Built with scientific rigor from Argentina. 🇦🇷
*Franco Luciano Carricondo - DNI 35.664.619*