Does Social Network Analysis Have a Place in Fraud Detection?

Does Social Network Analysis have a place in fraud detection and prevention?


Part I

Professionals estimate that around $300 billion is lost to public assistance fraud in the US annually – and half of that is believed to be stolen by organized crime groups.

In 2007, California’s Contra Costa County’s civil grand jury estimated child-care fraud costs county taxpayers $500 million annually. County officials agreed to study “data mining” systems in 2007 after reports were published that showed chronic fraud in federal, state and local public assistance programs by criminal enterprises. The data mining system will assign a numerical score to all welfare recipients that will alert investigators about suspicious people.

The system will use activities and characteristics of past welfare cheats to create a computer model that assigns a “risk score” to help identify new cheats. The new data mining system also has an advanced option called “social network analysis” (SNA). SNA helps investigators see relationships between people and assistance providers to create a relationship picture of suspicious people, associations, groups and behaviors.

Financial institutions are also exploring the benefits of SNA. According to Ellen Joyner-Roberson, Financial Services Marketing Manager at SAS: Social network analysis, also known as link analysis, is a powerful tool in understanding the structure of social and organizational networks that are often connected to criminal behavior. SNA maps and measures relationships and flows among people, groups, organizations, computers or other information/knowledge processing entities. The nodes in the network are the people and groups, while the links show relationships or flows between the nodes. Standard rules-based systems can't unearth "first-party fraud" and "bust-out fraud" where criminals establish accounts for the sole purpose of committing fraud.


A classic example is found within the credit card industry. TowerGroup, a Needham, Mass., research and analysis firm, projects that total card credit losses for issuers of U.S.-branded cards will peak at $55.6 billion in 2009. Rules-based systems are looking at more traditional types of risk, such as poor credit. With SNA, fraud-based risk can be seen by investigators, making it easier to uncover previously unknown relationships and conduct more effective investigations.

According to Joyner-Roberson, banks’ first concern is to know and authenticate the customer so they know with whom they are doing business. They must take a 360 degree view of their customer. Social network analysis, especially using more sophisticated analytics, can be used to find previously undetected fraud rings.