Machine learning for money laundering detection

Abstract

Money laundering nowadays has become more severe and gained attention from regulations all over the world. With the development of technology, machine learning and data mining techniques have been adopted in the anti-money laundering system. However, banks face difficulties when investigating cross-bank transactions due to the restrictions in information sharing between them. This project first studies the typologies of money laundering activities based on international organizations’ information and constructs indicators to quantify red flags for potential money laundering activities. From that, we propose an anti-money laundering framework comprising of single-bank and multi-bank systems by using machine learning, homomorphic encryption, and secure multi-party computation techniques. Future work can be done for the real-world implementation of such a framework.