Uber facial recognition is locking Indian drivers out of their accounts 


Uber checks {that a} driver’s face matches what the corporate has on file by a program referred to as “Actual-Time ID Test.” It was rolled out within the US in 2016, in India in 2017, after which in different markets. “This prevents fraud and protects drivers’ accounts from being compromised. It additionally protects riders by constructing one other layer of accountability into the app to make sure the precise particular person is behind the wheel,” Joe Sullivan, Uber’s chief safety officer, stated in a press release in 2017.

However the firm’s driver verification procedures are removed from seamless. Adnan Taqi, an Uber driver in Mumbai, bumped into bother with it when the app prompted him to take a selfie round nightfall. He was locked out for 48 hours, an enormous dent in his work schedule—he says he drives 18 hours straight, typically as a lot as 24 hours, to have the ability to make a dwelling. Days later, he took a selfie that locked him out of his account once more, this time for a complete week. That point, Taqi suspects, it got here all the way down to hair: “I hadn’t shaved for a couple of days and my hair had additionally grown out a bit,” he says. 

Greater than a dozen drivers interviewed for this story detailed cases of getting to search out higher lighting to keep away from being locked out of their Uber accounts. “At any time when Uber asks for a selfie within the evenings or at evening, I’ve needed to pull over and go underneath a streetlight to click on a transparent image—in any other case there are possibilities of getting rejected,” stated Santosh Kumar, an Uber driver from Hyderabad. 

Others have struggled with scratches on their cameras and low-budget smartphones. The issue isn’t distinctive to Uber. Drivers with Ola, which is backed by SoftBank, face comparable points. 

A few of these struggles might be defined by pure limitations in face recognition expertise. The software program begins by changing your face right into a set of factors, explains Jernej Kavka, an unbiased expertise guide with entry to Microsoft’s Face API, which is what Uber makes use of to energy Actual-Time ID Test. 

Adnan Taqi holds up his telephone within the driver’s seat of his automotive. Variations in lighting and facial hair have probably brought on him to lose entry to the app.


“With extreme facial hair, the factors change and it could not acknowledge the place the chin is,” Kavka says. The identical factor occurs when there may be low lighting or the telephone’s digicam doesn’t have a great distinction. “This makes it troublesome for the pc to detect edges,” he explains.

However the software program could also be particularly brittle in India. In December 2021, tech coverage researchers Smriti Parsheera (a fellow with the CyberBRICS mission) and Gaurav Jain (an economist with the Worldwide Finance Company) posted a preprint paper that audited 4 industrial facial processing instruments—Amazon’s Rekognition, Microsoft Azure’s Face, Face++, and FaceX—for his or her efficiency on Indian faces. When the software program was utilized to a database of 32,184 election candidates, Microsoft’s Face didn’t even detect the presence of a face in additional than 1,000 pictures, throwing an error fee of greater than 3%—the worst among the many 4. 

It might be that the Uber app is failing drivers as a result of its software program was not educated on a various vary of Indian faces, Parsheera says. However she says there could also be different points at play as effectively. “There might be a lot of different contributing components like lighting, angle, results of ageing, and so forth.,” she defined in writing. “However the lack of transparency surrounding the usage of such programs makes it laborious to offer a extra concrete clarification.” 


Please enter your comment!
Please enter your name here