Offline CBDC Fraud Detection

This project focuses on fraud detection for offline CBDC transactions using a hybrid architecture combining rule-based systems and machine learning models.

Main contributions include:

• Fraud-risk scoring
• Explainable AI (SHAP)
• Random Forest and LightGBM models
• Synthetic transaction generation
• Embedded deployment considerations

The work was developed using the OffSim-CBDC simulator.