Big data based fraud risk management at Alibaba

TLDR

Fraud risk has become pervasive with the rise of mobile internet and finance. The paper introduces Alibaba’s big‑data fraud risk management system and its AntBuckler product, aiming to detect and prevent diverse malicious behaviors for merchants and banks. Alibaba’s system uses real‑time big‑data processing, machine‑learning models, and the RAIN score engine to capture fraud signals, predict high‑risk users and transactions, and present risk scores and connections through a visualization UI.

Abstract

With development of mobile internet and finance, fraud risk comes in all shapes and sizes. This paper is to introduce the Fraud Risk Management at Alibaba under big data. Alibaba has built a fraud risk monitoring and management system based on real-time big data processing and intelligent risk models. It captures fraud signals directly from huge amount data of user behaviors and network, analyzes them in real-time using machine learning, and accurately predicts the bad users and transactions. To extend the fraud risk prevention ability to external customers, Alibaba also built up a big data based fraud prevention product called AntBuckler. AntBuckler aims to identify and prevent all flavors of malicious behaviors with flexibility and intelligence for online merchants and banks. By combining large amount data of Alibaba and customers', AntBuckler uses the RAIN score engine to quantify risk levels of users or transactions for fraud prevention. It also has a user-friendly visualization UI with risk scores, top reasons and fraud connections.