Fraudsters, Be Very Afraid

Fraudsters, Be Very Afraid

Fraudsters, Be Very Afraid

 

The fraud-fighting dream team of the future likely includes machines and humans. Picture this: your credit union’s fraud prevention system flags a series of transactions as high risk. The transactions, however, are for gas and groceries. Seems fairly normal, right? The system may have gotten it wrong. Upon closer inspection from a team of data scientists and analysts, though, you discover the fraud was all too real.

What gives machines + humans true dream-team potential is the unique skillsets they each bring to the table. Machines can be fully automated to perform complex tasks and predict behavior – eliminating the need for costly, time-consuming manual reviews. This is what is known as machine learning. Humans can then further analyze the machine outputs in light of their experiences with the wider fraud environment.

In true futuristic fashion, machine learning involves teaching computers to think for themselves. The computers identify patterns, solve problems and respond as programmed for certain scenarios. “The precise nature of machine learning helps ensure fraud detection is highly accurate,” said Ashley McAlpine, fraud prevention manager at CO-OP Financial Services. “Combined with a human review, flagged transactions can be quickly verified as fraudulent or not.”

Beyond checking if a machine’s fraud consensus is correct, humans add an extra layer of insight. Take the example outlined above, for instance. In this scenario, the human element confirmed the fraud and then took it a step further by looking at the bigger picture. This approach revealed a “fraud road trip” trend where criminals used stolen cards to make a long string of purchases within a few days. Machines identified what fraud was happening, while humans identified what it meant – another dream team challenge complete.

Already, companies are integrating machine learning into consumers’ day-to-day activities. Google’s search engine, for example, watches how consumers respond to results. With satisfactory findings, consumers will likely click one of the top hits. If, however, the consumer moves on to the next page, the program learns the results weren’t optimal and makes adjustments for the future. In a similar fashion, Netflix and Amazon use machine learning to make recommendations.

For credit unions, along with the fraud-fighting capabilities of machine learning comes the potential to positively impact member experiences. “The promise of machine learning is that transactions will be faster and more frictionless,” said John Buzzard, CO-OP’s industry fraud specialist. “This affords members a more comfortable transaction experience without unusual denials they don’t understand.”

Credit unions partnering with CO-OP Financial Services gain access to emerging machine learning technology. “It’s the way of the future,” said McAlpine. “Our credit unions should have the best-in-class fraud prevention technology backing them – and that includes machine learning.”

Transparency is key in a successful execution of machine learning. Complex, numeric results mean little to a member at the other end of a denied transaction. Machine learning solutions should deliver easy-to-understand results for credit unions and their members.

With any foray into machine learning, Buzzard says credit unions should:

  • Keep an open mind. Machine learning has been around since the 1950s. While the notion of teaching computers to think can be a little daunting, it is not a new one, and the technology has evolved leaps and bounds since its inception.
  • Be flexible. With any up-and-coming technology, new advancements are consistently on the horizon. The next iteration of machine learning is already in the works.
  • Maintain a collaborative network. Sharing the fraud trends machine learning uncovers can help benefit the credit union industry as a whole. By being proactive and collaborative, credit unions have a better chance of preventing widespread fraud.