Privacy-Preserving Federated Learning Engine for Credit Scoring
Decentralized credit default prediction across simulated bank silos with local model gradient training and differential privacy noise.
Project Overview
Trains machine learning models across decentralized financial institutions without sharing private customer records. Utilizes Federated Averaging (FedAvg) combined with Laplace differential privacy noise to protect training datasets against model inversion attacks.
Trains machine learning models across decentralized financial institutions without sharing private customer records. Utilizes Federated Averaging (FedAvg) combined with Laplace differential privacy noise to protect training datasets against model inversion attacks.
Decentralized credit default prediction across simulated bank silos with local model gradient training and differential privacy noise.
Core Project Objectives
Capture continuous analog/digital sensor readings with robust noise filtering and hardware calibration.