Facial recognition technology captures an individual’s unique facial data, extracting it to create a digital facial signature. What are the benefits and challenges of digital identities?
The Singapore government has announced plans to further develop the National Digital Identity (NDI) initiative by embracing AI and facial recognition technology to keep and maintain a national biometric database by 2025.
The eventual goal is to remove the hassle of dealing with passwords and physical IDs, and to expand the use of this technology beyond surveillance and security.
DigiconAsia sought out Fanglin Wang, Head of Artificial Intelligence, at ADVANCE.AI, to find out how the technology will play a key role in Singapore’s national digital identity programmes, what countries in this region can learn from implementations in other parts of the world, and the key issues and challenges surrounding the technology.
What technology innovations can we expect to be deployed for Singapore’s national digital identity programme, as part of its Smart Nation initiative?
Wang: Facial recognition will be a key part of Singapore’s national digital identity (NDI) programme, under its Smart Nation initiative. Facial recognition technology works by capturing an individual’s unique facial structure, features and skin tone, extracting it to create a digital facial signature. This is then compared via algorithms against a comprehensive public or private database, and finally determining if there is a match.
Already we see this technology being used widely at Changi Airport’s Terminal 4 and border immigration, and high traffic public areas like HDB, MRT, and bus interchanges for security. As part of the NDI initiative, the goal is for the government to keep and maintain a national biometric database by 2025.
By doing this, it removes the need for people to remember passwords or even identity card numbers and we will see broader use cases in areas like hospitality (e.g. hotel or theme park entry), food and beverage (personalized order history), supermarkets (payments and orders), building and event entry, and crowd control measures.
Are there other national or municipal digital identity implementations elsewhere in the world, and what can Singapore learn from them?
Wang: In China, facial recognition technology is a way of life. It scans people and monitors errant behavior such as jaywalkers or motorists who flout traffic rules or even people spending excessive amounts of time gaming. This is then used to build a “social credit system” which impacts your credit loan scores or ability to rent housing.
Commuters in the country
can use facial recognition technology to automatically authorize payments
instead of scanning a QR codes on their phones. It’s also used widely at
airports, railway stations, banks, schools, and shopping centers for
identification and payment authentication. Chinese people have accepted facial
recognition tech as a way of life and are comfortable with it.
In the US, social platforms like Facebook and Instagram are arguably the world’s largest facial recognition databases with millions of selfies and photos uploaded daily. Facial recognition has been heavily used in law enforcement and surveillance as well, but accuracy and false positives have been concerns.
Civil liberty and the right to privacy are part of the constitution in US, and to certain extent the UK and West in general. Countries like Singapore, meanwhile, have to find their own way and balance along that scale.
How have AI and facial recognition technology been deployed within banking, financial services, retail, e-commerce and e-payment ecosystems?
Wang: AI and facial recognition technology is being deployed quite widely in the banking, financial services, retail, e-commerce, and e-payment sectors. It can really drive business efficiency and accelerate digital onboarding, fraud prevention, and process automation for enterprises. This is especially relevant now with the crippling impact of Covid-19 on businesses worldwide, from large MNCs to smaller SMEs, which are having to deal with limited face-to-face interaction.
Demand is high across these sectors. We’ve seen demand for our flagship product, ADVANCE Guardian, actually increase 400% year-on-year. We went from 100 to 400 clients in a single year (through January 2020). Over the same period, our monthly API call volume has increased by 350% across six markets. So facial recognition technology is in high demand.
In Indonesia we helped Home Credit, one of the country’s top online underwriters, with alternative credit modelling, ultimately lowering the risk performance of their portfolios by providing modelling with more predictive power. In Indonesia, Danamart, one of the country’s top peer-to-peer (P2P) platforms, used our facial comparison and “liveness detection” solution to accelerate their customer verification process. Our credit scoring and customized scoring also improved their credit risk analysis, both in speed and accuracy.
In India, we helped CASHe, one of the country’s top digital lending apps, with automating national identity documentation, with 99% accuracy. The ultimate goal is to help them digitalize the entire customer onboarding process.
Please share some pros and cons of facial recognition technology, including weaknesses as it exists today, and possible future developments that could overcome these weaknesses.
Wang: It’s important to recognize that facial recognition technology has already been part of daily life for almost a decade. It’s widely used in the algorithms in our smartphone cameras when taking selfies to identify faces in the shot, or automatically group and tag photos. As mentioned, social networks can be considered the world’s largest facial recognition databases today when you consider the amount of selfies and photos uploaded to these platforms. Of course, mobile banking applications use facial identification to verify customer log-ins, and you also see this technology in use at airports and border checkpoints.
As with any technology, there are pros and cons. Based on where facial recognition technology is today, some of the pros include it being fast, convenient, and contactless. It removes the need to remember passwords or other digital tokens. It is very hard to falsify or defraud, and it is more accurate than human eyes. Good facial recognition technology and algorithms can be more than 99 percent accurate, while human verification is less than 98 percent. Best of all, it can run 24/7 day and night, and in a place like Singapore where human resources and process optimization are really critical, this technology can really be a game-changer.
Naturally, there are some drawbacks. For example, collected data must always be kept safe and secure with a trusted guardian and safekeeper. Data must only be used for specific and stringent purposes, such as for law enforcement. A balance must be maintained between convenience and data privacy and security. It’s important for governments and the private sector to provide consumer education and comfort when it is initially introduced (i.e. at Changi Airport in Singapore). Finally, facial recognition technology must be able to prove a very high degree of accuracy before it’s deployed widely to prevent misidentification, and extensive testing must be conducted to understand its limitations and prove accuracy before the technology is deployed.
What are some key ethical, privacy, and security concerns surrounding the use of facial recognition? How much should governments regulate for it to be both sufficiently safe and efficient?
Wang: Governments should regulate facial recognition technology to ensure it is used properly and not abused. For example, use should be for stringent and specific purposes such as law enforcement or border immigration control, and data must be securely stored and protected.
Extensive testing should be conducted to understand the limitations and accuracy of the technology before it is deployed more widely.
Even as Singapore pushes ahead with its Smart Nation agenda, it is also one of the leaders in promoting right and ethical use of AI. The Singapore government proposed updates to its AI Governance Framework at Davos 2020 in January.
Broadly speaking, this requires that human involvement must be paired with AI technology to ensure accountable decision making. Moreover, AI decision-making must always be explainable, transparent, and fair.
At the end of the day, every country needs to find the right balance and comfort between convenience and privacy. Every country needs their own regulation and enforced by the government to prevent bad actors and abuse.