News Stay informed about the latest enterprise technology news and product updates.

Big data takes a bite out of credit card bust-out fraud

Credit card bust-out fraud, the power of YouTube and conflict prediction: The Data Mill reports.

Fraud is a headache, no doubt about it, but some types of fraud are a bigger headache for financial institutions than others. Take credit card bust-out fraud. That's when fraudsters create a bunch of identities, open up new accounts and buy stuff to drive up the lines of credit before maxing out the cards and disappearing.

The Data Mill"If you're really crafty, you'll write fake checks to clear off the [credit card] balance because [financial institutions] will zero your balance before they clear a check, and then you max out the cards a second time," Ari Gesher, software engineer for Palo Alto, Calif.-based Palantir Technologies Inc., said during a webinar on data, crime and conflict hosted by O'Reilly Media.

Credit card bust-out fraud is hard to spot because new accounts are, well, new. There are no patterns of behavior yet, which makes spotting deviations impossible, Gesher said. So all those statistical models built to determine if a credit card has been stolen? They don't work here.

Of course, there's more than one way to catch a thief. For credit card bust-out fraud, the solution requires a horizontal view of the business and a symbiotic human-machine relationship, according to Gesher. Integrating siloed data from account details and transactions records to payment information, call center and Web server data gives financial institutions a chance to cluster seemingly unrelated accounts together for a different perspective.

"Are unrelated accounts used in the same store minutes apart from each other, accessed with the same IP or caller ID data? Are the account details or payment methods similar?" Gesher said. These could be signs of fraud, but are practically invisible to the human eye in real time.

Machines, however, can sift through the vast amounts of data to find those details, cluster accounts together and score them. Humans then dig into high-risk clusters to determine if something foul is afoot, an important step because sometimes what looks like fraud isn't fraud.

One of Palantir's customers spotted unusual behavior that looked like the mother lode of credit card bust-out fraud, Gesher said. A closer examination revealed that the clustered accounts were linked together by a single phone number, which turned out to be an outgoing line for a large company with something like 100,000 employees. "False positives are a real issue here," he cautioned.

YouTube armed

The Carter Center, the human rights organization founded by former President Jimmy Carter, is up in arms over social media, in a manner of speaking. The Atlanta-based nonprofit is "curating YouTube videos of declarations by Syrian arms groups," Helena Puig Larrauri, a consultant who works with nongovernmental and United Nations agencies, said during the webcast.

Videos are watched by analysts and tagged "with a very comprehensive list of categories," she said. Networks are analyzed to see how different groups relate to and interact with each other. Analysts at The Carter Center are also keeping tabs on the major players on Twitter and Facebook.

"One of the interesting things comes out of Twitter analysis; they're able to deduce where funding and alliances are coming from," Puig Larrauri said.

Previously on
The Data Mill

International Institute for Analytics' predictions for 2014

Add semantic analysis to ward off big data/bad analytics syndrome

Big data tech and Bitcoin at MIT VC Conference

Big data goes to war

Could big data coupled with machine learning help foster world peace -- or at least help the United Nations mitigate the horror of war? Chris Perry, a policy analyst for the New York-based International Peace Institute, thinks so. "If you can collect enough actionable intelligence information in advance of things like conflict or mass atrocities, you can marshal resources to prevent things from escalating," he said.

Perry has been looking at ways to "operationalize things like machine learning as part of an early warning toolkit." He helped build a sample model and tested it against 2012 data. The results proved promising, he said. "We're hoping to develop this into a usable model … over the next year."

Social media on fire

By now, most businesses understand they need a Twitter account as a core facet of their communications programs. But disaster management expert Jeannie Stamberger, director of innovation for the Kampala-based ResilientAfrica Network, warns that organizations shouldn't be lulled into thinking social media is a cheap and simple way to interact with constituents. "Doing it wrong can be costly," she said.

For some public agencies, that means maintaining two distinct Twitter feeds. The Los Angeles Fire Department uses one account for incident alerts. A second account is more of a PR channel, providing a place for information, conversation and even a little personality. "You can chat with firefighters, know them by name," Stamberger said. For either channel to work, the city must invest in it, even allocating firefighters time for tweeting, she said.

Welcome to The Data Mill, a weekly column devoted to all things data. Heard something newsy (or gossipy)? Email me or find me on Twitter at @TT_Nicole.

Dig Deeper on Enterprise business intelligence software and big data

Join the conversation

1 comment

Send me notifications when other members comment.

Please create a username to comment.

Tell us how your company is capitalizing -- or not! -- on social media.