Facial payments in public transport: 6 takeouts from Moscow’s subway pilot

Only a few cities around the world use facial payments in public transport. It’s still a new technology and little is known about convenience ansd security of that kind of biometric systems.

So I was enthusiastic to dive into the subject last month when writing a story for TU-Automotive about Moscow subway’s Face Pay. Since the end of 2020, the technology has being been tested with the subway’s employees and contractors. In August 2021, testing with a limited group of passengers had also been started.

Today, regular operations of Face Pay must be launched, according to the mayor’s earlier pledge. My greetings to the subway’s passengers and the Face Pay creators who, by now, have won TechFest’s two nominations: Artificial Intelligence and Transport Infrastructure Champion.

Moscow isn’t a world-first city to implement a large-scale facial payment system in its public transit network. In some Chinese cities, similar big projects had been launched as early as 2019. However, Chinese authorities reluctantly share any takeouts from their experience. This is why watching the Moscow pilot is especially interesting.

I. Is facial recognition always accurate?

Let me share a personal anecdote on this issue. A friend of mine is a big enthusiast of the government’s national biometric system. Theoretically, it can be used as a single pass to the large bunch of financial and public services. So, he recently told with glee that he’d registered. He gloomily added: “My smartphone is the only device that recognizes me.”

To make sure it’s getting a reliable solution, Moscow subway had held a competition between three developers VisionLabs, Ntechlab and Tevian to find a solution with the best rate of identification errors. To us, the key outcome is error rates are different. In other words, errors do happen.

I’d asked the city’s press service if two-step identification is used to increase robustness of identification. The response was that, “currently, it’s not needed but it can be demanded later when the number of users grow.”

However, it remains unclear yet how the issue translates into the practice of facial payments.

II. Is facial payment faster than payment cards?

It takes 1.5 seconds for the Face Pay to process one passenger. Meanwhile, it’s 0.3 to 0.8 seconds for the city’s transport card and 1.0 to 1.5 for a bank card or smartphone. Thus, facial payments is not as fast as other contactless payments. Not yet, according the city’s press service, because fine-tuning of facial recognition technology is not complete so speed is expected to further grow.

In airports, though, facial recognition checkups were proved to save boarding time around nine minutes per wide-body aircraft.

III. Do facial payment systems store images of people’s faces?

A properly built system doesn’t. Instead, it stores hashes (cryptographic codes) of images which don’t contain necessary data for image reconstruction, according to experts. Devil’s in the details, however, as there is no sure way for a usual customer to find out if the system is a “properly built” one.

IV. What should a customer be concerned about?

What are people afraid of? Typical concern is that some scoundrel uses an image stolen from, say, a social network profile to ride for free. Personally, that was my concern as well. However, can you imagine that somebody suggests a picture of another person to a turnslite cam while in a danger of been caught by other passengers and subway personnel? Nonsense!

A malicious one wouldn’t even approach a turnslite. In fact, social engineering is far the biggest threat. It aims at your bank account binded to the public transport payment server.

V. Why people in China and Russia are not as enthusiastic about facial payments as people in EU and the US?

Analysts say that’s because of concerns over possible breaches of people’s facial biometric data. Back in 2019, nearly 80% of Chinese people were concerned about it. Their concerns proved to be real by 2021 when Chinese authorities prosecuted a criminal group that hacked a government-run facial recognition system to fake tax invoices worth $76M. In the Western countries, stronger protection of personal data results in higher consumers’ trust in the technology.

In a recent survey, 74% to 75% of Russians indicated they trust nobody with their biometry and don’t plan to use it for commuting purposes. On the contrary, 2020’s report by FIDO Alliance notes that more than 50 percent of consumers in the UK say they will use biometrics for payments.

VI. Why EU and the US lag behind China or Russia in facial payments in public transport?

Analysts say that technological leadership has nothing to do with that. In China and Russia, the city authorities’ interest in the technological development because it helps to collect people’s personal data. Biometric data can be later used in two ways, for public space surveillance and trading consumer-related data. In the Western countries, legislative protection of personal data strongly limit both surveillance and personal data trade.

Surveillance is not all bad. Take Moscow’s case: before facial payments were introduced, the city tested the facial recognition technology in the subway’s surveillance system. In nine months, it helped to find 1800 people wanted by the police including suspects and 170 lost people. Particularly, it helped to find 39 lost children.

However, it also can be used to suppress political protest. So public control over the authorities is necessary. In the words of Margrethe Vestager, executive vice-president for a Europe fit for the Digital Age: “On artificial intelligence, trust is a must, not a nice to have.”

I’m grateful to the press service of Moscow’s department of transport, analysts Natalia Pchelovodova of J’son&Partners and Nikolay Legkodimov of KPMG, and Vladimir Pedanov, CEO at Autovisor, who helped me with this story by sharing their expertise.