Graph-Powered Analytics and Machine Learning with TigerGraph

Chapter 5. Detecting Fraud and Money Laundering Patterns

In this chapter, we take on the serious problem of fraud and money laundering. Fraud is typically conducted by one or more parties as a multistep process. Sometimes, the only way to distinguish fraud or money laundering from legitimate activity is to detect a characteristic or unusual pattern of activity. Modeling the activity and relationships with a graph enables us to detect suspicious activity by searching for those patterns along with checking for their frequency.

After completing this chapter, you should be able to:

• Describe coordinated activity among multiple parties in terms of a graph pattern

• Use a multihop or iterated single-hop graph traversal to perform a deep search

• Describe bidirectional search and its advantages

• Understand the use of timestamps to find a time sequence

Goal: Detect Financial Crimes

Financial institutions are responsible for averting criminal money flows through the economic infrastructure. According to The Financial Action Task Force (FATF), illicit funds amount to 3.6% of global GDP.¹ A well-known criminal activity is money laundering, or disguising the origin of money earned through illicit means. According to the FATF, 2.7% of global GDP is laundered per year. Banks are legally obligated to investigate their clients’ payment behavior and report any suspicious activities.

Other types of financial fraud include identity theft, where someone uses another person’s accounts without ...