AI Security Labs

⚡ Quick Start

# Setup and run complete ML01 lab
git clone <repository-url>
cd ai-security-labs
make setup && make ml01-full

🎯 Features

🔬 OWASP ML01: Input Manipulation Lab

• FGSM Attack Implementation: Fast Gradient Sign Method with batch processing
• Multi-Method Defense System: Gradient norm + confidence analysis with ROC curves
• Professional CNN Model: Enhanced SimpleCNN with validation and debugging
• Comprehensive Evaluation: Attack success rates, defense metrics, visualization
• OWASP Compliance: Follows OWASP ML Security Top 10 guidelines

🚀 Commands

# Complete setup
make setup

# Run ML01 lab components
make ml01-train     # Train OWASP ML01 model
make ml01-attack    # Generate adversarial examples 
make ml01-defense   # Evaluate defenses
make ml01-full      # Complete OWASP ML01 pipeline

# Development workflow
make lint           # Code quality checks
make test           # Run test suite
make clean          # Clean generated files
make monitor        # Monitor system performance

� Core Components

SimpleCNN Model (src/owasp/ml01_input_manipulation/model.py)

from src.owasp.ml01_input_manipulation.model import SimpleCNN

model = SimpleCNN()
predictions, confidence = model.predict(data)
gradients = model.get_gradients(data, labels)

FGSM Attacker (src/owasp/ml01_input_manipulation/attack_fgsm.py)

from src.owasp.ml01_input_manipulation.attack_fgsm import FGSMAttacker

attacker = FGSMAttacker(model, device)
adversarial_data, attack_info = attacker.fgsm_attack(data, labels, epsilon=0.25)

Adversarial Defense (src/owasp/ml01_input_manipulation/defense_fgsm.py)

from src.owasp.ml01_input_manipulation.defense_fgsm import AdversarialDefense

defense = AdversarialDefense(model, device)
is_adversarial, detection_info = defense.detect_adversarial_combined(data, labels)

📦 Requirements

• Python: 3.8+ with virtual environment support
• PyTorch: 1.9+ with optional CUDA acceleration
• Development Tools: Included in requirements-dev.txt

🤝 Contributing

1. Fork and clone repository
2. Setup development environment: make setup
3. Create feature branch
4. Implement changes with tests
5. Run quality checks: make lint test
6. Submit pull request

📄 License

MIT License - see LICENSE file for details.

Author: Gopesh Khandelwal [email protected]

🔒 Build secure, robust ML systems with hands-on adversarial learning!