Naman-Tulsyan/rift-money-muling

# 💳 Money Muling Detection Engine ### 🚨 AI + Graph Intelligence System for Detecting Fraud Rings in Financial Transactions --- ## 🏆 Hackathon Project Submission This project is an **end-to-end fraud detection platform** designed to identify **money mule accounts, coordinated fraud rings, and suspicious transaction behavior** using a hybrid approach that combines: - 🧠 **Graph Analytics** - 🤖 **Machine Learning** - ⚙️ **Rule-Based AML Detection** - 📊 **Interactive Visual Investigation Dashboard** Our system detects **fraud networks BEFORE financial damage occurs**, making it highly suitable for real-world deployment in banking, fintech, and digital payment ecosystems. --- # 🌍 Problem Statement Financial fraud involving **money mule accounts** is rapidly increasing, especially in high-volume digital payment systems like UPI. Traditional fraud systems suffer from major limitations: ❌ Focus only on individual transactions ❌ Detect fraud after loss occurs ❌ Cannot identify coordinated fraud rings ❌ High false positives (flagging legitimate merchants) There is a strong need for a system that can: ✔ Detect fraud networks early # Money Muling Detection Engine (Graph + ML) End-to-end platform for detecting potential money mule accounts and coordinated fraud rings from transaction data. It ships as: - A FastAPI backend that ingests a transactions CSV, builds a directed transaction graph, detects suspicious ring patterns, computes explainable suspicion scores, and optionally blends them with an ML model. - A Next.js dashboard that lets you upload data, visualize the transaction graph, inspect detected rings, review per-account risk, and download a JSON report. ## What this repo actually does ### Detection pipeline Given transactions (sender → receiver edges with amount + timestamp), the backend runs: 1. CSV validation and normalization (Pydantic model parsing for timestamps/amounts) 2. Graph construction (NetworkX `MultiDiGraph`) 3. Pattern detectors - Cycl …