Biometrics is smart, but AI is smarter. Here's why

AI and biometrics can work together to develop effective and reliable security models.

In the digital age, security is one of the primary concerns of any organization. Every organization has realized that data is a major resource. Hence, organizations deploy advanced security mechanisms to safeguard business data. However, various industry giants such as Google, Facebook, Cathay Pacific, Exactis, and many more have witnessed major data breaches in the past few years. These data breaches have exposed vital data of millions of customers. Therefore, businesses are constantly looking for better alternatives to traditional security models.

Biometrics such as fingerprint and iris scans are being utilized for authenticating employees at the workplace and identifying smartphone owners. Such biometrics can be implemented in organizations to authorize data access for confidential data. Biometrics can be used along with traditional passwords or PIN numbers for multi-factor authentication. Additionally, the adoption of AI will help develop data-driven security protocols. Hence, clubbing AI and biometrics together will lead to the creation of dynamic security models.

How do biometrics work?

Biometrics identifies and authenticates individuals based on their physical or behavioral characteristics. This method can be more secure compared to traditional methods such as passwords and PIN numbers that can be easily hacked. Biometric systems utilize sensors that convert biometric traits such as fingerprints, face, iris, and voice of a person to an electrical signal. The type of biometric sensor is chosen based on its application. For instance, a high definition camera will be required for facial recognition. Biometric systems can be divided into various types:

• Fingerprint recognition: Fingerprint scanners collect an image of a person’s fingerprint and record its features such as whorls, loops, and arches. Additionally, fingerprint scanners also analyze outlines of edges, furrows, and minutiae of a fingerprint. The scanned fingerprint image is then verified against a previously stored set of fingerprints.
• Iris recognition: Iris scanners analyze features of a person’s iris such as rings, freckles, and furrows that are situated around the pupil. Iris scanners contain a video camera that can scan irises through contact lenses and glasses.
• DNA matching: DNA matching uses a physical sample of an individual such as blood, hair, or saliva to confirm their identity. DNA matching is widely used in forensic investigations due to its unparalleled accuracy.
• Ear acoustic authentication: Size and shape of every human ear are unique. The size and shape of human ears help in collecting sound waves and routing them with the help of ear canal. Special earphones with a microphone can capture sound waves reflected inside the ear canal. The reflected sound waves captured by the microphone are used for identification of an individual.

Along with these types, there are several other types of biometrics such as hand geometry, palm vein recognition, signature recognition, and odor recognition. Different kinds of biometrics are used in various scenarios such as forensic investigations, airport security, access control, and banking.

How is AI smarter than biometrics?

Although biometrics are trusted for accurate authentication mechanisms, advanced technologies such as artificial intelligence can deceive biometric systems. Some researchers were able to fool biometric systems with the help of AI-generated synthetic fingerprints. Researchers detailed a major flaw in the biometrics used in devices such as smartphones. Instead of using a full fingerprint, multiple devices allow users to submit several fingerprint scans. Biometric systems analyze partial fingerprint scans to find matches with saved partial fingerprints. Such partial fingerprints can generate inaccurate results.

Researchers developed a machine-learning algorithm to create synthetic fingerprints. These AI-generated synthetic fingerprints can easily fool current biometric systems. With the help of fingerprint data, AI-generated fakes can become more unique and accurate. Such fingerprints can be used for brute force attacks that can test every generated fingerprint until the target device is unlocked. Hence, several devices can be at risk of brute force attacks with the help of synthetic fingerprints.

Artificial intelligence and machine learning have displayed untold potential in cybersecurity. Likewise, AI and biometrics can be clubbed to create a secure authentication mechanism. The combination of AI and biometrics can help develop authentication systems that can protect devices against cyber attacks and prevent fraudulent activities.

How can AI and biometrics work together?

AI and biometrics can work together in the following manner:

Keystroke dynamics

Keystroke dynamics is the method of identifying and authenticating people based on their typing patterns. Keystroke dynamics can identify individuals with the help of their speed, dwell time, and flight time. Dwell time measures the duration of a key being pressed and flight time is the interval between releasing a key and clicking another key. The time required for a person to search for the right key and the time spent while pressing the key can be calculated together to authenticate individuals. Keystroke dynamics can be used for multi-factor authentication along with PIN numbers or password. However, keystroke dynamics have not become mainstream due to its inaccuracies. AI and biometrics can be clubbed to make keystroke dynamics more precise and a reliable typing pattern recognition technology. AI systems can track information about how individuals type and the time interval between two keys for the most frequently used keys to identify individuals. Also, AI can learn from user profiles that are built over time as users continue typing.

Facial recognition

Facial recognition is a popular feature among several smartphones and on social networks such as Facebook. However, facial recognition can be easily tricked. For instance, Samsung Galaxy S10’s face unlock can be fooled with the help of a video or picture of the owner. In some cases, the face unlock feature can also be fooled by the face of an owner’s sibling. Such incidents have also happened in the past due to inaccuracies in 2D facial recognition. Also, several law enforcement agencies use facial recognition with the help of surveillance cameras to identify criminals in public spaces. But, the entire face of a criminal may not be visible in public places. Hence, law enforcement agencies are looking for better alternatives.

Facial recognition can be more effective with the help of machine learning. AI learns from millions of images and utilizes 3D biometrics to successfully authenticate an individual’s face. AI systems can also use predictive modeling to analyze the effects of aging on human faces. For this purpose, AI analyzes pictures of old people to recreate younger images of those people. With the help of large volumes of available facial data, AI and biometrics can together create more precise authentication models.

Voice recognition

Several smart home devices such as Google Home and Amazon Alexa use voice and speech recognition for multiple tasks like answering queries, ordering products, and playing music. However, these devices cannot authenticate users before doing any of the given tasks. Additionally, Google has incorporated voice recognition-based ‘Smart Lock’ in several Android devices. But, this unlocking method can be inconvenient to use as it can fail while identifying the voice of a user and is unable to perform in noisy situations.

The implementation of AI in voice recognition can train the biometric systems using millions of voice samples of different users. AI and biometrics like voice recognition can evaluate a person’s biometric voice signature by analyzing their voice patterns such as speed, accent, tone, and pitch. Such biometrics can be quick and authenticate individuals precisely. Such AI-powered voice recognition can be used in workplaces for authentication and attendance purposes.

Gait detection

Gait detection is a method that authenticates individuals based on the way they walk. This authentication technique has been researched for decades but it was never conventionally adopted. However, gait detection can be a viable authentication solution with the help of AI. Researchers at The University of Manchester obtained an accuracy of 99.3% in gait detection using AI. AI-powered gait detection analyzes a person’s steps with the help of floor sensors. With the help of AI, gait detection can be deployed for airport security and diagnosis of several medical conditions.

The combination of AI and biometrics can precisely verify the identity individuals based on their physiological and behavioral traits. However, their adoption and implementation are limited due to the lack of commercial applications. Hence, business leaders can introduce AI-based biometrics for their workplaces and customers to offer user-friendly and secure authentication protocols. With this approach, AI-powered biometrics may become mainstream soon.