BoostFex
IEEE Access implementation: bank‑transaction anomaly detection via Gradient‑Boosted Federated Learning (privacy‑preserving).
Overview
BoostFex hosts the official implementation of “Detection of Bank Transaction Anomalies using Gradient Boosted Federated Learning” (IEEE Access). It explores how gradient‑boosted models can be trained across distributed parties while keeping sensitive transaction data local.
Problem
Banks need anomaly detection that is accurate and fast — but centralizing transaction data is costly and creates privacy/compliance risk. The goal: learn from multiple data owners without moving raw data, while still achieving strong detection performance.
Approach
Implemented a federated learning workflow for gradient‑boosted models, separating client‑side training from server‑side aggregation. Built repeatable experiments to compare configurations and validate privacy‑aware training at scale. Designed the codebase to be modular: data handling, training, federation logic, and evaluation are cleanly separated for extension.
Impact
A research‑grade, reproducible codebase aligned to a peer‑reviewed publication — ideal for demonstrating applied ML rigor, privacy‑by‑design thinking, and distributed systems engineering.