Social network analysis for fraud detection in payments

Social network analysis for fraud detection is becoming an increasingly important and useful tool. As payment fraud continues to rise, particularly in the card-not-present (CNP) space, this approach is proving an effective fraud detection technique. Despite its name, social network analysis has nothing to do with Facebook or Twitter. But what is social network analysis and what are its benefits for dealing with fraud?

What is Social Network Analysis?

Traditional methods for detecting fraud involve a silo approach. Rather than looking across products or channels, fraud prevention has relied on risk score based analysis. Banks would set rules and computers would check to decide whether or not to allow a transaction based on these. The problem is this means trying to predict what constitutes fraudulent behaviour.

A range of modern data analysis tools are improving fraud efforts and social network analysis (SNA) is one of the most important. SNA is a form of link analysis that allows fraud officers to see patterns across multiple channels. At its simplest, it’s just a way of representing different actors (nodes) and their relationships (ties) by way of a graph.

“Unlike other analytical techniques like statistics that are based on the notion of independence of subjects, SNA can provide useful insight into large datasets along network, spatial and time dimensions based on the interconnectedness of the subjects being analysed,” notes a CGI Group paper on the topic.

Often referred to link analysis, it is a data mining technique that reveals the structure and content of a body of information. It represents this as a set of interconnected and linked objects. Kelvin Chan and Jay Leibowitz outlined some of the key methodology for SNA in their 2006 paper “The synergy of social network analysis and knowledge mapping: A case study”.

SNA and payment fraud detection

For the financial industry, SNA has a key part to play in detecting fraud as to a large extent it’s being committed by organised networks. What the analysis does is put some context around the content of the data. In payments and banking, the social networks consist of accounts, card numbers, customers, phone numbers, email addresses and so on. Connections (ties) between these various nodes can be analysed to determine patterns that could indicate fraudulent behaviour.

SNA has a particularly strong use case for first-party fraud. A lot of fraud detection is based on finding patterns in card transactions, but there is no pattern to the transaction side of first-party fraud, which often involves synthetic schemes based on the creation of a fake ID to obtain credit cards, overdrafts or loans. Patterns often exist, however, in the networks of individuals perpetrating these crimes.

Because of this, SNA is proving to be a key tool. Rather than looking at the risk of an individual transaction, the social network analysis approach offers a far broader picture to the fraud officer, placing it among the wider context of the network. SNA tackles the root of the proposed fraud before it gets to the transaction where the money is lost.

For example, when groups of individuals are creating fake identities for loan applications, SNA can be used to flag suspicious behaviour by showing the connections of things like addresses, phone numbers, email addresses.

SNA is not a cure-all for fraud, but it is proving to be exceptionally useful for certain variants as part of a broader fraud prevention strategy.


Written by David Divitt