While globalization has enabled easy cross-border movement and transactions, fraudulent activities have, unfortunately, followed suit. The growing demand for a multi-layered and improved security system—whether for border control, online activities, or public security—is driving the need for reliable global authentication. As security demands skyrocket, innovative ideas are emerging to resolve many of today’s urgent issues.
Biometric identity capture is currently the most promising solution to secure the movements of every individual—online and in the physical world—and to prevent fraud of any kind. The field is working on a further headway: behavioral biometric technology. The way the behavioral aspect complements biometrics could cater for safer, more reliable, and faster identification.
Sophisticated and reliable: Traditional biometric technology
With traditional biometric technology, adopters can identify any registered person through their biometric data such as the fingerprint, the iris, or, most commonly, facial features. Used as a rapid tool to access smartphones or a digital banking app instead of typing a passcode, the technology has immensely improved our user experience when interacting with technology.
Next to the high adoption for digital identity authorization, biometric technology has been setting foot in law enforcement. Police departments and other agencies in Europe, the US, and elsewhere rely on public and private photo databases to identify suspects. The technology also helps to conduct real-time surveillance of public spaces and recognize criminal suspects in a group of people. When it comes to controlling building entrances and exits, the technology assists in regulating visitor access and identifying potential security risks (such as robberies, terrorism, and unauthorized passage).
Public facial recognition technology is commonly used at country entry ports and in public transport systems. In China, for example, passengers use their faces instead of physical tickets. In the European Union, facial recognition technology is used at borders and at airports to verify people’s identities for visa and asylum applications. Several other countries, including Israel, are planning to use the technology for border security and at checkpoints.
So far, public agencies have always been able to rely on the latest research on detection technology and the evaluation of different algorithms. The US National Institute of Standards and Technology (NIST) provides an accurate list of which recognition technologies, such as fingerprint or facial recognition, are mature enough for offering reliable and trustworthy results. For example, the latest industry standard for facial recognition is the NIST FRVT (Facial recognition vendor technology). The NIST FpVTE (Fingerprint vendor technology evaluation) evaluates promising fingerprint recognition services and what to consider when deciding on a provider.
Revolutionary and immature: Behavioral biometric technology
In contrast to traditional biometrics, behavioral biometric approaches are “younger” and less standardized. This makes an unbiased comparison without an industry benchmark such as the NIST FRVT or NIST FpVTE much more difficult. The standard distinguishing feature of behavioral biometrics is that they provide a confidence value regarding a particular identity. Fingerprints or faces can give a definite “yes” or “no”, although confidence scores are also widely adopted for traditional biometrics.
Nevertheless, behavior-based approaches such as movements, keystrokes, phone swipes can serve practical purposes. For example, they can become a second factor for rapid identity verification or a supporting factor for biometric identification. In addition, in cases where actual identification features are obscured (e.g., use of facial masks or gloves), behavioral biometrics can provide helpful clues when authorities need to identify someone reliably.
The use cases of behavioral biometrics
There are several promising applications in the field of behavioral biometrics. For computer-based identity verification, there are solutions that allow identification based on keystrokes—the frequency and patterns of which prove to be individual enough to recognize identity. Due to the nature of typing, the models can also get better because they can continuously monitor and analyze keystroke data. Software developers tend to also customize confidence thresholds depending on the use case.
However, in some cases, the reliability of this behavioral biometric factor is limited to the circumstances. On a different keyboard, individual patterns may differ, and physical conditions like carpal tunnel syndrome or arthritis may affect unique abilities. The lack of benchmarks makes it difficult to compare different providers’ trained algorithms in these cases, providing room for false marketing claims.
Image analysis for image recognition can provide more data for behavioral research. Gait and posture biometrics are rapidly becoming useful tools, even if they do not yet match the accuracy and robustness of traditional biometric approaches. Nevertheless, the system’s ability to see a moving person and suggest an identity based on gait is a good preliminary filter for public safety concerns, such as access control.
Let’s say that a camera observes a person approaching a turnstile from a distance and selects possible candidates from a watch list. Once close enough, the system can verify the identity using facial recognition. Analyzing aisles helps narrow down possible identities and saves time and resources in cases where watch lists include hundreds or thousands of people (in large buildings or at airports). With the embedding of platforms like NVidia Jetson that can turn ordinary cameras into smart devices, pre-processing performance improves.
Police forces in countries that comply with the EU General Data Protection Regulation (GDPR) also use biometric data on gait and movement to identify potentially dangerous situations in real time. For example, if the cameras see two people in a violent discussion, they can notify security staff and promptly send a police unit to resolve the problem. The system still needs visual confirmation from a supervisor but collecting sufficient data can advance detection.
In addition, the system can obscure faces to protect personal data while still recognizing meaningful interactions. Police officers can still carry out direct identification, depending on the legal environment they’re subject to.
Behavioral biometrics are in their infancy but offer high potential
There are other approaches to behavioral biometrics, such as analyzing the heartbeat or other physical phenomena. Still, this information must be uniquely individual and easy to use for the method to be effective. The heartbeat may be promising once wearables can measure details such as the electrical impulses generated by the heart comparably and accurately.
However, they too can change in the event of cardiac arrest. Therefore, gait and posture are the most promising starting points, as they do not require the active participation of the person observed. When identifying keystrokes, typing is part of the process and, therefore, not intrusive.
Behavioral technology is not mature enough yet to guarantee reliable and safe usage for achieving public and personal safety. However, looking at the current strive for optimization in the industry, it is only a matter of time before the technology is ready to evolve from secondary use to more widespread applications.