Can Machine Learning Stop Uber Drivers From Getting Scammed?

July 3, 2018 - 5 minutes read

Uber is having trouble taking care of a particular crime problem. The San Francisco-based developer of the eponymous ride-hailing app is no stranger to scams; recently, it’s been utilizing machine learning, a subset of AI, to deal with con artists.

But can it apply this technology to thwart this new trend targeting drivers?

Built on Lies

In early June, an Uber driver was dropping off a passenger in L.A. when he decided to accept a new trip in the Beverly Hills area. That’s when he received a call from a number with the San Francisco 415 area code. “They tell me to pull over and cancel the trip,” explained the driver, who asked not to be identified during an interview with CNET. “They tell me I need to verify my account. They sound very professional and very poised.”

The caller told the driver, who had only been working for Uber for two months, that the passenger on his next trip was disabled, so Uber wanted them to get an SUV instead. Believing this, the driver canceled the trip. After this, the caller requested the driver’s email, Uber account password, and two-factor authentication code that he’d receive through text. The driver divulged all of this information.

He explains: “When you’re a new driver and you get a call like that, it’s hard to put on your logic hat and say, ‘I’m just going to guess you’re a scam artist.’ But recounting the story now, it’s very clear to me that that was a red flag.” At the end of the week, his $653.88 in earnings had been transferred from his Uber account to an unfamiliar debit card. They also locked him out of his own account.

After checking with Uber, it turned out that the caller had been the same person who initiated the driver’s Beverly Hills trip. Uber agreed to credit the driver the full amount stolen back.

Stopping Other Scams With Machine Learning

Uber’s used to dealing with scams. In the past, con artists have signed up as fake drivers just to collect fees on fake cancellations. Other times, thieves cheated passengers by offering fake discount rides on the Internet. Uber developed machine learning just to weed these fraudsters out. And it seems to be working — since implementing some measures, the company reports stopping hundreds of frauds.

Their machine learning tools use the data from the app’s millions of users to track down the crook. According to Uber’s data science manager, Ting Chen, the tools take more than 600 signals into account while detecting red flags. In one instance, it was the parameter of altitude that led to the scammers’ downfall. “They created some very weird altitudes,” explains Chen. “These trips were actually flying in the sky.”

But this time around is different: it’s aimed at taking advantage of the drivers, not the passengers or the company.

A Growing Problem

The scam has already hit thousands of drivers, adding up to millions of dollars stolen, according to a lawsuit filed by the U.S. Attorney’s Office in New York’s federal court. Uber says it is working with law enforcement to capture these criminals, and that greater security measures are now being enforced for drivers’ accounts.

“Overall, the effectiveness of these scams has dramatically reduced since last year as a result of additional fraud rules instituted by our security teams,” an Uber spokeswoman explains. The company also states that it has sent out notification warnings about this scam to its drivers.

Yet, all three interviewed by CNET reported receiving nothing of the sort. According to Harry Campbell, a driver based in L.A. who runs the Rideshare Guy blog, Uber could be doing more: “The volume has increased on this, as opposed to decreased. Uber hasn’t done anything to warn drivers.”

This problem is perplexing to address. Certainly, Uber is investigating how its current machine learning tools could help solve it. What do you think the company’s best course of action is?

Tags: , , , , , , , , , , , , ,