The integration of biometrics and AI
We’ve all been entertained by surreal examples of artificial intelligence (AI) in sci-fi movies: from smart homes that trigger a lockdown at the slightest hint of danger, to facial security scans that only grant access to invited guests. But these technologies are no longer confined to the world of fiction — they are becoming a reality for consumers.
Innovations like AI and the Internet of Things (IoT) are increasing the inter-connectivity between people and devices. As a result, there is a growing need to ensure people can safely use digital platforms without the fear that their data will become compromised. This is where biometrics and AI come in.
While biometric techniques measure a person’s physical characteristics to verify their identity through physiological traits like fingerprints and irises, AI is playing a pivotal role in making such security systems even smarter. Given AI’s ability to collect and process huge amounts of data at speed, it can establish patterns and detect threats when certain patterns are disrupted; this could be when a person’s facial features don’t match with any in a given database.
Applications
Payment card fraud is probably the first thing that comes to most people’s minds when we consider the best application of AI-driven biometrics for digital security. Supporting this, a 2018 study by Experian found that almost three-quarters of the businesses surveyed cited fraud as a growing concern over the previous 12 months1. This is not surprising given the huge costs incurred as a result of fraud — the Nilson Report2 suggests that worldwide losses from credit card fraud could exceed $31 billion by 2020.
But it’s important to remember that the application of AI in biometrics isn’t just limited to online fraud. In fact, AI solutions are poised to deliver major improvements to security, not to mention time and cost savings, across a broad range of markets. From protecting personal information stored in our smartphones, to identifying threats in crowded public areas like airports, there are many ways that AI is being used to protect consumers and businesses.
Perfect companion
The core appeal of AI systems is they are able to work without human intervention. These technologies have a high level of autonomy and ability to adapt to new situations based on previous experiences. This makes AI a perfect companion for biometric authentication, which involves processing huge stores of information and constantly responding to new security threats.
AI neural networks: these provide the frameworks for multiple algorithms to work together and process data inputs, underpinning AI solutions like those used in biometric authentication.
Together, these two characteristics mean that AI can play a central role in the development of next-generation biometric solutions. For example, AI can be trained to mitigate inconsistencies and overcome obstructions, in order to acquire a match based on a biometric data sample like a fingerprint or eye scan. So while in the past recognition systems have been limited by constraints like the kinds of images that could be used, AI is enhancing this functionality to offer more effective scanning.
In terms of the best applications of combined biometric and AI systems, facial recognition stands out. Face identification systems have become established as one of the most common biometric authentications that consumers regularly come across, with providers ranging from phone manufacturers to airport security teams and social media platforms all adopting this technology. The commercial possibilities are vast, and as a result the facial recognition market is growing at a remarkable pace — projected to be worth an impressive $7.7 billion in 2022, up more than $3 billion from 20173.
At the same time, facial recognition systems performance continues to improve. Last year, NIST (the US National Institute of Standards and Technology) reported that between 2010 and 2018 facial recognition technology on average got 20 times better at searching a database to find a matching photograph4: in 2010 5% of algorithms failed to match a face in a given database; last year this figure had fallen to just 0.2%.
Smartphone protection is among the most evident and practical applications of this technology, with the Apple iPhone X Face ID system providing clear benefits for users in securing their device: Apple estimates that the chance of a random face unlocking your phone is about one in a million. Samsung is also making steady progress in this field of developing holistic biometric solutions. Its smartphones now offer iris scanning security which recognises the unique patterns in a person’s irises (see https://www.samsung.com/global/galaxy/galaxy-s8/security/). These patterns are virtually impossible to replicate, making iris authentication one of the safest ways to keep phones locked.
Protecting the home
Biometric authentication is also increasingly being used to screen who comes into our houses as part of AI-enabled home security systems. AI systems are significantly better at recognising faces than humans and can determine instantly whether a person at the door is within the database of people who are permitted entry.
Based on this, the ‘smart doorbell’ market is growing. For example Tuya (https://en.tuya.com) recently unveiled its new AI Video Doorbell, which uses facial recognition to identify each member of a family using a photograph — thereafter allowing them access to the property. Tuya claims that the smart home platform is able to respond to a wide variety of scenarios, including detecting strangers who might be lurking around the property and trying to gain access. Cleverly, the system is able to automatically turn on lights and music to ward off a potential intruder.
Google, too, has long been toying with AI-powered home security systems, and launched its face-scanning smart doorbell — Nest Hello — into the UK market last year. The doorbell has an integrated camera and speaker system, which is designed to connect to other smart home devices. On arrival, a guest will trigger a response that will send an alert to the homeowner, allowing them to see who is at the door.
Online banking
AI-powered biometric solutions have also begun to proliferate in the e-commerce and online banking sector. Faced with the task of overseeing millions of transactions on a daily basis, banks are constantly on the lookout for new innovative ways to secure their customers’ personal data and finances, and are deploying AI widely to fight fraud and related cyber-crime.
Take Mastercard as an example. It is responsible for monitoring more than 2 billion cards in over 210 countries and territories, and according to its own figures, processes 165 million transactions an hour. In response, Mastercard is now using machine learning (ML) algorithms to examine each transaction. Its ML and AI solutions are able to provide key information about each individual purchase, including what the item in question is and where it was purchased. Based on a history of previous spending patterns, the technology is then able to detect whether a certain transaction looks suspicious.
Building on this, Mastercard is now increasingly leveraging AI solutions in conjunction with biometrics. Fingerprint, iris and facial recognition tools are helping the bank to verify the identity of card users and ensure that only legitimate transactions are processed. This is done through the Mastercard Identify Check5, which encourages mobile shoppers to authenticate a purchase through either a finger scan, or by showing their face to the device so it can perform a facial recognition check. Upon verification, the funds are released.
Enhancing the user experience
Ultimately, AI-driven biometrics can offer greater security and convenience. The recognition and authentication processes are typically instantaneous, minimising the need for people to rely on their memory to retrieve passwords and PIN numbers. Furthermore, it removes human error from the process; a hacker can no longer guess a person’s password, which are often predictable and used across multiple platforms. This combination of AI and biometrics also inhibits criminals from acquiring the personally identifiable information needed to commit fraud: the data used to verify someone’s identity cannot be stolen or shared, ensuring that legitimate cardholders have the confidence to spend money online.
The response from consumers to this has been positive, particularly among the younger age groups. Research from Vocalink shows that ‘Millennials’ have thrown their support behind AI and biometric authentication, confirming that consumer sentiment is driving demand for these technologies. In 2017, 28% of those aged 18–35 had already used fingerprint technologies to verify payments, while 35% considered this form of technology to be the most secure method of verification6. We can expect both of these figures to increase notably over the coming years.
Future developments
So what can we expect from the AI-biometrics collaboration in the future? While great strides have already been made in this field, and despite all the buzz that biometric recognition technologies have generated, it’s important to remember that we’ve only just scratched the surface of this technology’s capability. Every day, the capabilities of AI and ML are advancing.
Over the coming years, some of the more realistic uses of AI in biometrics we are likely to see will include typing and writing behaviour biometrics. It’s known that a person’s handwriting is unique to them and so can act as a form of identification that is extremely difficult to falsify. However, subtle differences in handwriting can go unnoticed without sophisticated means of scrutinising the information. AI can be used to solve this obstacle. The technology can take into account timing and pressure, as well as the slant of different letters and the order of operations; in essence, all the subtle writing habits that an individual demonstrates.
These analysis techniques can also be applied to keystrokes. AI can learn a person’s typing habits and determine if someone is trying to impersonate them. Identifying factors like typing accuracy, speed, hand-dominance and pressure can all be analysed to create a means of drawing attention to anomalies to the established patterns. Equally, one major benefit of AI in biometrics is that it allows for constant monitoring. So while a person’s typing behaviour might change over time, AI tools are able to continuously study the individual’s habits and account for gradual changes, allowing the technology to build a broad typing profile of a user. This in turn allows the system to verify that the person editing a document or filling out information requests is consistently who they say they are.
In saying this, it’s very difficult to predict how this field will change in the coming years, particularly considering the speed at which this technology is advancing. But as organisations prioritise the security of their customers, there are undoubtedly more sophisticated AI-fueled biometric safeguards on the horizon.
Nikolas Kairinos is the chief executive officer and founder of Soffos, the world’s first AI-powered KnowledgeBot. He also Founder & President at Fountech.ventures.
