Moneycontrol PRO

The startup behind DigiYatra reveals how it perfected the tool’s facial recognition algorithm

Avinash Komireddy, Founder and CEO of Dataevolve, the start-up that designed the DigiYatra ecosystem, claims the facial recognition algorithm has a success rate of 99%

February 07, 2023 / 08:08 AM IST
The six-year-old Hyderabad-based start-up has developed the digital architecture of DigiYatra, which is currently operational in Delhi, Bengaluru, Varanasi and Hyderabad (Representative Image)

The six-year-old Hyderabad-based start-up has developed the digital architecture of DigiYatra, which is currently operational in Delhi, Bengaluru, Varanasi and Hyderabad (Representative Image)

The facial recognition algorithm of DigiYatra, the biometric and contact-less security clearance system at airports, has been trained using six lakh unique faces, concentrating on the Indian demography. The algorithm has a success rate of 99 per cent, said Avinash Komireddi, Founder and CEO of Dataevolve.

The six-year-old Hyderabad-based start-up has developed the digital architecture of DigiYatra, which is currently operational in Delhi, Bengaluru, Varanasi and Hyderabad. Although the system has not been officially launched in Hyderabad, it has been operational there for a while now.

DigiYatra is being heralded by the government as the future of air travel, as it provides touchless passenger validation through facial recognition, which saves time at various touchpoints like entry to the airport, security hold area, and boarding area with no intervention from security personnel. The system authenticates one’s facial biometrics against their Aadhaar.

However, there has been criticism regarding the application of facial recognition technology, particularly in terms of privacy, surveillance, discrimination and digital divide. It has often been pointed out that the country still lacks a data protection law to protect against misuse of sensitive data, such as biometrics, and so on.

Komireddi’s statement of DigiYatra’s facial recognition algorithm being trained on a six-lakh India-specific dataset is significant as it has often been pointed out that “facial recognition is inaccurate, especially for people of colour (which includes Indians) and women".

“It’s a homegrown algorithm, with forks from public repositories. And we created a dataset, especially on the Indian demography; that was a key focus,” he said.  A fork is a copy of an existing software project that has modifications on top of it.

Komireddi also claimed that the algorithm has a 99 percent success rate, and provided to demonstrate an example for Moneycontrol. One of Dataevolve’s developers, in his early 20s, matched his facial biometrics against his Aadhaar photo, that was taken when he was a teenager. It took a few seconds for the DigiYatra architecture to successfully verify the photo of the bearded developer, against his over-a-decade-old Aadhaar picture.

Currently, Komireddi said, almost 1,500 people use DigiYatra in New Delhi per day on average, 800-1,000 people in Bengaluru, 600 in Hyderabad, and around 200 in Varanasi.

The system and how it works

Firstly, passengers have to sign up on the DigiYatra app available on Android and iOS. “So, we first take his facial biometrics, and match it against the person’s Aadhaar.”

The facial recognition authentication process takes place on the Amazon Web Services (AWS) cloud platform. “That’s the only touch point where your data is going into AWS, because doing it (the verification) on the phone is not very practical,” he said, adding that no data is stored on AWS, or with the start-up themselves.

“If it is a match then we issue a credential from the blockchain,” Komireddi said. Dataevolve uses the Aries Hyperledger blockchain platform, which is designed for solutions focused on creating, transmitting and storing verifiable digital credentials.

“The credentials from the blockchain will ensure that the passenger does not have to go and authenticate personally over and over again with Aadhaar,” he said, adding that the credential is saved on the user’s wallet.

“The matched face is always on the blockchain, and every time it is retrieved by airports for verification purposes, the blockchain ensures that the credential is not tampered with,” he added. The data is stored with airports for 24 hours, following which it is deleted.

Checks and balances

Dataevolve’s CEO concurs with the concerns regarding privacy and surveillance associated with technologies such as facial recognition, but assured that DigiYatra’s architecture has been designed with those concerns in mind.

Komireddy explained that the entire architecture of DigiYatra was reviewed multiple times by the technical working committee of DigiYatra Foundation, whose members include experts from IIT, industry experts such as those responsible for building United Payments Interface (UPI), and so on.

Recently, the Standards and Testing Quality Centre (STQC), an attached office of the Ministry of Electronics and Information Technology, undertook a four-month audit process. The STQC-mandated audit is set to take place every year.

“The DigiYatra Foundation was very conscious about the fact that the users should have control over their data,” he said.

Current issues

“Scaling on the blockchain network is very difficult,” Komireddi said, while pointing at the recent negative reviews that the DigiYatra app has been receiving. Users have complained a lot about the facial authentication process not going through, credentials not being generated, and so on.

He said that when the user’s wallet, which stores the credentials, sends it to the blockchain for verification, the blockchain is unable to respond in time.

“Once you have, say 1,000 or 1,500 people registered at the same time, that's where no matter how big a server you throw at that computer, it's unable to respond back to the mobile device immediately,” he said, adding that one has to find ways to navigate around it.

Initially, Dataevolve also faced issues in integrating with various airports.

“Each airport has its own vendor, and we had to standardise the way we communicate with the airport,” he said, adding that, currently, the process has smoothened after they developed a process over six-seven months.

What’s next?

Apart from getting ready to scale the ecosystem for the entire country, Dataevolve is also planning to provide additional information to the user, such as the necessary travel time if one uses DigiYatra.

“Instead of reaching the airport like 3-4 hours before lift-off, one can actually get to understand the rush hour, and so on, and plan their airport travel accordingly,” he said.

They also plan to provide airports software development kits so that users can download their DigiYatra credentials from the DigiYatra app and import it to the airport’s app.

“So that if someone is a frequent traveler from Bengaluru, he or she can port the credentials to the Bengaluru airport app, and share the credentials right away,” he said.

Going forward, he also plans to take this ecosystem to other countries, such as Dubai and Australia.

Invite your friends and family to sign up for MC Tech 3, our daily newsletter that breaks down the biggest tech and startup stories of the day

Aihik Sur covers tech policy, drones, space tech among other beats at Moneycontrol
first published: Feb 6, 2023 05:06 pm