Intelligent Anti-Money Laundering System * Shijia Gao, Dongming Xu, Huaiqing Wang, and Yingfeng Wang Abstract-Criminal elements in today's technology-driven society are using every means available at their disposal to launder the proceeds from their illegal activities. While many anti-money laundering solutions have been in place for some time within the financial community, they cannot adapt to the ever-changing risk and methods in relation to money laundering. In order for a more adaptive, intelligent and flexible solution for anti-money laundering, the intelligent agent technology is applied in this research. Intelligent agents with their properties of autonomy, reactivity and proactivity are well suited for money laundering prevention controls. Several types of agents are proposed and a novel and open multi-agent architecture is presented for anti-money laundering. A prototype system for money laundering detection is also developed to demonstrate the advances of the proposed system architecture and business value. Index Terms-Anti-money laundering, Artificial intelligence application, Intelligent agents. I. INTRODUCTION S! ince the mid-1980s, money laundering (ML) has been increasingly recognized as a significant global problem, with serious economic and social ramifications [1]. Today, ML has become a key funding mechanism for international religious extremism and drug trafficking, and curtailing these illegal activities has become an important focus of governments as part of their ongoing wars on terrorism and drug abuse. Following the terrorist activity of September 11, 2001, there has been an increased focus in the United States and across the globe on the prevention of ML and terrorist financing. Increasingly, anti-money laundering (AML) systems are being implemented to combat ML. However, the traditional rule-based solutions suffer from a number of drawbacks, such as ineffective thresholds, high false positive problem, lack of pattern recognition function, and insufficient data Shijia Gao is with the School of Business, University of Queensland, Australida (e-malCbl.G. busjness.ug.edu.au Dongming Xu is with the School of Business, University of Queensland, Australia (e-mail: D.Xubusiness uu edu.au) Huaiqing Wang is with the Department of Information Systems, City University of Hong Kong, Hong Kong (SAR) (Corresponding Author. Phone: 852-2788-8491; fax: 852-2788-8535; e-mail: iswaneg_ncty u.edu. t dhk' Yingfeng Wang is with the Department of Information Systems, City University of Hong Kong, Hong Kong (SAR) (e-mail: 500952650Xstud,ent.citxyu.edu .hk). This research is supported by a strategic research grant (No. 7001805) from the City University of Hong Kong. processing capability. In this research, we apply intelligent agent technology to ML prevention controls by taking advantage of agent's autonomy, reactivity, proactivity, and social ability. The organization of this paper is as follows. Next section briefly reviews the relevant literature on ML, AML, AML systems, and intelligent agent. Section III presents the architecture, development, and operation of a multi-agent-based AML system. The final section addresses some conclusions as well as the future work. II. BACKGROUND A. Money Laundering andAnti-Money Laundering Money laundering (ML) is a term usually used to describe the ways in which criminals process illegal or "dirty" money derived from the proceeds of any illegal activity (e.g. the proceeds of drug-dealing, human trafficking, fraud, theft or tax evasion) through a succession of transfers and deals until the source of illegally acquired funds is obscured and the money takes on the appearance of legitimate or "clean" funds or assets [2]. ML is a diverse and often complex process that need not involve cash transactions. ML basically involves three independent steps that can occur simultaneously [3]: * Placement - the process of transferring the proceeds from illegal activities into the financial system in such a manner as to avoid detection by financial institutions and government authorities. * Layering - the process of generating a series or layers of transactions to distance the proceeds from their illegal source and obscure the audit trail. * Integration - the unnoticed reinsertion of successfully laundered, untraceable proceeds into an economy. The International Monetary Fund (IMF) estimates that the aggregate size of ML in the world could be somewhere between 2 and 5 percent of global gross domestic product (GDP), equivalent to approximately US$590 billion to US$1.5 trillion annually. According to Celent Communications, the amount of illicit funds traveling through ML channels is estimated to reach over US$926 billion worldwide by the end of 2005, and grow at an annual rate of 2.7%. However, those are just estimates - the full magnitude of the problem is still not known with any certainty. Recent years have witnessed a growing number of highly publicized money laundering scandals involving major 1-4244-0318-9/06/$20.00 ©2006 IEEE 851 Authorized licensed use limited to: George Mason University. Downloaded on December 22, 2009 at 14:10 from IEEE Xplore. Restrictions apply.