Liveness Detection 101

Digital identity fraud continues to be a top concern for businesses and consumers. An ever-growing range of sophisticated tools empower fraudsters to steal personal information and account access credentials, compromising system security. Fastest growing among these is the use of artificial intelligence (AI) to create synthetic identities or even mimic legitimate users’ faces and voices.

One of the most powerful tools in a fraudster’s arsenal is the use of Generative Adversarial Networks (GANs). GANs can create hyper-realistic images, videos, and voices, making it possible for criminals to produce highly convincing digital impersonations that can mimic legitimate users with alarming accuracy, commonly known as deepfakes. These deepfakes can easily fool traditional biometric systems, while leveraging personal details that have found their way into the Large Language Models (LLM) AI draws from to correctly respond to any knowledge based authentication they may encounter, allowing fraudsters to bypass even the most secure identity verification processes.

At the same time, consumers are increasing their acceptance of biometrics for secure access, placing their transactions directly in the crosshairs of these advanced fraud tactics. Research conducted by PYMNTS reported that, “most modern consumers now see biometric tools, which smartphones conveniently provide, as the most secure way to authenticate a transaction.”

The threat is real: Thomson Reuters reports that, “Over 80% of all new account fraud can be attributed to synthetic identity fraud.” This statistic is staggering, but understandable when you consider the ease in which new technology not only allows for the creation of a non-existent person, but can also make that person appear completely real, both visually, and in action. The cost is also staggering. According a report from The Insurance Information Institute, four of the top five types of identity fraud in 2023 were opening new credit card accounts, opening new bank accounts, applying for and receiving government benefits, and taking out personal and business loans.

However, all is not lost. Incorporating liveness detection technology into biometric onboarding and authentication ensures that the data being presented is not only accurate but also comes from a real, valid document and a live, real person. By incorporating liveness detection, businesses can effectively neutralize the threats posed by AI-generated deepfakes and other synthetic identities.

What is Liveness Detection?

At its simplest, liveness detection is exactly what it sounds like, detecting that an image seen by a camera, or a voice heard by a microphone is a live person. In a “presentation attack” fraudsters use masks, photos, videos, voice recordings, and deepfakes to pass themselves off as a legitimate account owner. Liveness detection leverages a number of different techniques to distinguish real people trying to open or access an account from criminals using these kinds of attacks.

According to Biometric Update, “Biometric liveness refers to the use of computer vision technology to detect the genuine presence of a living user, rather than a representation such as a photograph or a mask…Typically associated with facial recognition, liveness can also be applied to voice recognition to distinguish present speakers from audio recordings, and finger or palm biometrics such as by detecting blood flow, and even iris recognition.“

How Does Liveness Detection Work?

Liveness detection has evolved far beyond the early checks of looking for a blink or smile to prove that a person is alive. Today, advanced liveness detection techniques leverage sophisticated AI algorithms and computer vision technologies to analyze biometric data at a level of detail that is imperceptible to the human eye. These technologies go beyond surface-level assessments to examine the intricacies of how light interacts with skin, the micro-expressions that occur naturally on a living face, and the subtle variations in voice patterns that cannot be replicated by recordings or synthetic voices.

For instance, in facial biometrics, advanced liveness detection might analyze the sub-surface scattering of light in the skin, which can reveal minute differences in blood flow—something that a photo, video, mask, or deepfake cannot mimic. Similarly, for voice biometrics, the system might assess the dynamic range of vocal tones, the timing of breath patterns, and the physical characteristics of how sound waves are produced by the human vocal cords, all of which are nearly impossible to fake with current AI technology. Also, liveness detection techniques aren’t limited to detecting whether a person is “live”. Some advanced techniques, including several patented algorithms developed by Daon Labs, can help to ensure that documents being presented for identity verification are actual IDs by analyzing the image to determine if the substrate, ink, and other elements of the ID match the known construction of the document and ensuring that the image does, or does not, match patterns from previous submissions.

Not only does liveness detection look for small imperfections, it also detects excessive perfection, something often associated with synthetically generated data. It’s been said that humans are perfect in their imperfection. This is never more true than in the ability to differentiate an actual person from a synthetic representation, but with liveness detection, it’s at a level imperceptible to the human eye.

And since liveness detection is built on AI, it leverages machine learning to constantly evolve and improve, which means that every time a new type of attack is discovered, the algorithm can be trained to recognize the new threat, and any similar to it, in the future.  This continuous improvement process ensures that liveness detection technology provides businesses with a robust, adaptive defense that is vital to staying one step ahead of fraudsters, even when the technology they use is accelerating at an ever more rapid pace.

Types of Liveness Detection

When discussing liveness detection, the various techniques fall into two categories: active and passive. Each has strengths and weaknesses; both are effective for preventing many of the common attacks that are deployed in an attempt to defeat biometric security.

Active Liveness Detection

The basic differentiator for active detection is that it requires the user to knowingly complete a task. These tasks could include head movements, blinking, or even something more advanced like holding their ID in front of their face or repeating a random word or phrase spoken by an agent or IVR. AI algorithms track these activities and combine that data with biometric data to determine whether the subject is a real person or synthetic representation.

Passive Liveness Detection

Passive liveness detection operates without any scripted action on the part of the user and is, more often than not, happening without the user being aware, operating in the background during both identity verification and authentication processes. Due to its invisible nature, passive liveness is faster and creates significantly less friction than active liveness detection. It is also much hard for criminals to defeat, since they can’t develop a work around for something they can’t see.

Passive liveness detection uses AI neural networks to analyze the user’s face or voice during the process of capturing their biometric. Within seconds, the algorithms assess qualities such as colors, skin textures, shadows, and image and audio artifacts, cadence, pitch, and tone, ensuring the input is from a live user.

Benefits of Liveness Detection

While biometric factors provide the highest level of security and user experience when compared with other factors, without liveness detection, they still create significant fraud risk. Liveness detection eliminates the majority of these risks helping to create the optimum solution for protecting your data and your customers’ identities.

Eliminate Fake Accounts

The two fundamental questions that underlie the identity verification processes are, “are you a real person?” and “are you who you say you are?”. While the process of scanning an ID and matching it to a selfie is designed to answer those questions, it is not without risk of fraud. Forged IDs and manipulated images have been used to fool security methods for decades, and their availability and realism continues to increase at alarming rates. Computers are now able to fabricate realistic and accurate fake IDs and images of non-existent people, that most viewers would be certain are real, in a matter of seconds. Advanced liveness detection eliminates these concerns, helping to create a truly secure, reliable onboarding process that keeps fraudsters from accessing your organization.

Better Secure Existing Accounts

Biometrics greatly enhance the security of authentication by adding a strong layer to multi-factor identity protection. They are resistant to phishing and other social engineering attacks, can’t be stolen like passwords, and eliminate the common customer frustration of remembering login details. However, biometrics can still be vulnerable to spoofing attempts—photos, videos, masks, or even deepfakes. Fortunately, advanced liveness detection stops these attacks by ensuring only the real account holder can access their account, providing a necessary level of additional security.

Maintain Regulatory Compliance

For organizations operating in heavily regulated sectors, compliance is not just a legal requirement—it’s a foundational element of their operational integrity. Regulations such as the General Data Protection Regulation (GDPR) in Europe, the California Consumer Privacy Act (CCPA) in the United States, and global Anti-Money Laundering (AML) and Know Your Customer (KYC) laws mandate rigorous standards for how personal data is collected, stored, and used.

A large part of meeting these requirements lies in accurately verifying the identity of customers. With the ability of liveness detection to shore up the biometric identity verification process, institutions beholden to these regulations can ensure compliance. Furthermore, because biometric verification and authentication, with liveness detection, is the most secure method of identity access, it minimizes the risk of data breaches which can meet with strict fines under some of these rules.

Build a Foundation of Customer Trust

As news cycles continue to be filled with stories of the latest data breach or hijacked identity, customers are becoming increasingly wary of the security being employed to protect their data and their identity. What used to barely be an afterthought has become a key decision factor in choosing who to do business with. At the same time, these consumers are only willing to go so far when trading security for simplicity. While biometric verification and authentication provide the perfect balance between security and user experience, they are not immune to the news cycle. Consumers have heard and read the stories about accounts being access with deepfakes, and many have fooled their own technology with an image. These consumers need to know that your technology stack has the tools in place not only provide easy of use, but also to defend against every attack.

Save Money

The bottom line is that fraud costs money and it’s not just the amount directly stolen by criminals. A report by Alloy found that nearly 60% of its respondent banks, fintechs, and credit unions in the U.S. and UK lost over $500,000 in direct fraud losses in 2023, with 25% of respondents losing over $1 million. These figures don’t account for the costs of recovery, regulatory fines, lost customers, and reputational damage. A study commissioned by Forrester Consulting reported that the actual cost of fraud was 4.41 times the lost transaction value for financial services and lending in the U.S. and 3.90 times in EMEA.

In today’s rapidly evolving digital landscape, the stakes have never been higher. Staying ahead of fraudsters is a necessity for maintaining customer relationships, ensuring compliance, and safeguarding your organization’s reputation. Biometric security is the best way to protect your organization and liveness detection is key to fortifying those processes against the most sophisticated threats.

Daon integrates liveness detection, including a number of proprietary and patented algorithms, into all of our biometric identity verification and authentication solutions. We can assist you in creating a tailored solution for your security architecture that positions you for success now and into the future.