# Catch Me If You Can: Semi-supervised Graph Learning for Spotting Money Laundering

**Authors:** Md. Rezaul Karim, Felix Hermsen, Sisay Adugna Chala, Paola de Perthuis, Avikarsha Mandal  
**arXiv:** [2302.11880](https://arxiv.org/abs/2302.11880)  
**Date:** 2023-02-23  
**Summary page:** https://www.summarizepaper.com/en/paper/2302.11880v1/

## Key points

- Money laundering is a serious problem where criminals move illegal money to untraceable destinations and integrate them into legitimate financial systems.
- Identifying such activities accurately and reliably is crucial in enforcing anti-money laundering (AML) measures.
- Existing approaches to detecting money laundering from graphs of money transfers between accounts of a bank have yielded lower detection accuracy, making them less reliable.
- In the paper "Catch Me If You Can: Semi-supervised Graph Learning for Spotting Money Laundering," the authors propose a new approach that employs semi-supervised graph learning techniques on graphs of financial transactions to identify nodes involved in potential money laundering.
- The proposed approach models the transactions in the form of multipartite graphs to detect the complete flow of money from source to destination.
- Experimental results suggest that their approach can spot money laundering from real and synthetic transaction graphs with higher accuracy than existing methods.
- This could be useful for law enforcement agencies and financial institutions in preventing illicit activities and enforcing AML measures more effectively.

## Summary

Money laundering is a serious problem where criminals use financial services to move large amounts of illegal money to untraceable destinations and integrate them into legitimate financial systems. Identifying such activities accurately and reliably is crucial in enforcing anti-money laundering (AML) measures. However, despite tremendous efforts, only a small fraction of illicit activities are prevented. Existing approaches to detecting money laundering from graphs of money transfers between accounts of a bank have yielded lower detection accuracy, making them less reliable. In this paper titled "Catch Me If You Can: Semi-supervised Graph Learning for Spotting Money Laundering," Md. Rezaul Karim, Felix Hermsen, Sisay Adugna Chala, Paola de Perthuis, and Avikarsha Mandal propose a new approach that employs semi-supervised graph learning techniques on graphs of financial transactions to identify nodes involved in potential money laundering. The authors note that existing approaches do not take into consideration that money laundering involves high-volume flows of funds through chains of bank accounts. The proposed approach models the transactions in the form of multipartite graphs to detect the complete flow of money from source to destination. Experimental results suggest that their approach can spot money laundering from real and synthetic transaction graphs with higher accuracy than existing methods. This could be useful for law enforcement agencies and financial institutions in preventing illicit activities and enforcing AML measures more effectively. Overall, this paper presents an innovative solution for identifying potential cases of money laundering using semi-supervised graph learning techniques on financial transaction graphs which could help improve detection accuracy and reliability when it comes to enforcing AML measures.