Select your localized edition:

Close ×

More Ways to Connect

Discover one of our 28 local entrepreneurial communities »

Be the first to know as we launch in new countries and markets around the globe.

Interested in bringing MIT Technology Review to your local market?

MIT Technology ReviewMIT Technology Review - logo

 

Unsupported browser: Your browser does not meet modern web standards. See how it scores »

{ action.text }

Flagging Fraud
But while computers have grown powerful enough to apply evolutionary prin­ciples to all sorts of problems, the “haystacks” have been multiplying at an even more dramatic rate. Consider consumer fraud. Credit card companies estimate that $.07 per $100 charged to credit cards is lost to fraud, costing the industry more than $1 billion per year in the United States alone. Yet writing traditional software to identify fraudulent charges remains phenomenally difficult. Why? Because the people perpetrating the fraud are experts at modifying their behavior to evade detection. It’s simply not possible to write a program that anticipates every possible scam.

But evolutionary algorithms can at least make computerized fraud detection more likely to succeed, argue the artificial-intelligence researchers who founded New York City-based Searchspace. The company sells a variety of programs that split up the haystack by looking for aberrant activity within precisely defined slices of existing account data, says Michael Recce, Searchspace’s chief scientist. The software uses tools dubbed “sentinels,” programmed with fraud detection rules. Multiple charges to the same debit card at a single store on a single day, for example, might automatically raise a red flag.

But the person racking up these purchases may simply be a forgetful Christmas shopper, not a thief. So the sentinels weigh in a variety of factors, such as a person’s prior activity at that store, in order to avoid “false positives” and flag only accounts that human experts would agree are suspect. Says Recce, “You can set the fitness criteria in a way that delivers both minimal fraud loss and minimal good-customer loss.”

Searchspace routinely hosts “pilots,” essentially software bake-offs that pit its algorithms against potential clients’ existing fraud detection systems. Participants bring in blind samples of historical data to see if Searchspace’s sentinels plant red flags in all the right places. Invariably, says Recce, the sentinels turn up not only the preflagged accounts but also a few more miscreants lurking in the background noise. “I don’t think there’s been one of those presentations where we haven’t had to pause things for a moment so that an executive could go step out to make a quick phone call,” says Recce, smiling.

0 comments about this story. Start the discussion »

Tagged: Biomedicine

Reprints and Permissions | Send feedback to the editor

From the Archives

Close

Introducing MIT Technology Review Insider.

Already a Magazine subscriber?

You're automatically an Insider. It's easy to activate or upgrade your account.

Activate Your Account

Become an Insider

It's the new way to subscribe. Get even more of the tech news, research, and discoveries you crave.

Sign Up

Learn More

Find out why MIT Technology Review Insider is for you and explore your options.

Show Me