Fighting fraud with graph analytics

Teradata chief analytics officer Bill Franks on how social network analysis can uncover fraud

Using graph analytics, also known as social network analysis, to analyse people and their contacts has broken up fraud rings according to Teradata chief analytics officer Bill Franks.

He gave the example of a US government agency which gave funds to families who were poor to pay for child day care while the parents worked.

“They started using graph analysis and broke up a $30-million-a-year fraud ring.”

The agency analysed social networks and used data mining to determine if individuals who used the service were likely to commit fraud.

It worked out that some people were making up names for children who didn’t exist and claiming they were enrolled in a child care centre.

Banks are using graph analysis for fraud prevention purposes.

“Banks will look for money moving between accounts in unusual ways. When they identify that a given owner of an account is trouble they can cause graph analysis to work back how that money went into that account,” he said.

Because fraudsters tend to deal with other fraudsters graph analytics can also identify them, he said.

The analysis has also being used to fight cyber crime.

For example, Franks said if people were trying to ping the Teradata website, and the traffic was mostly coming from Russia, it could potentially determine from the IP addresses that it might be coming from a town with a lot of hackers.

“If people on social media are communicating back and forth and referencing other people, there are those links too,” he said.

Graph analytics has also been used by telcos to analyse calling patterns and churn models to see which customers were most likely to cancel their account.

“Say I am tightly connected to a group of friends and family and we are all on the same carrier,” he said.

“As I quit to go to another carrier, everyone’s risk goes up a little bit. If three people close to me switch to that same other carrier everyone has the risk exponentially.”

He said that for years, telcos have been reaching out to networks of people that have partially defected to take strong actions to stop them from defecting.

Follow Hamish Barwick on Twitter: @HamishBarwick

Follow Computerworld Australia on Twitter: @ComputerworldAU, or take part in the Computerworld conversation on LinkedIn: Computerworld Australia

Join the Computerworld newsletter!

Error: Please check your email address.

Tags Graph analyticssocial network analysisfraud

More about BillTwitter

Show Comments