Liveness Check

Anyone opening a bank account online today or signing up for certain services often has to show their face via camera. The system then checks not only whether the face matches an ID document – it also checks whether there is a real human being behind it. That is exactly what a liveness check is: an automatic authenticity check designed to detect whether the person is really sitting live in front of the camera right now. Without this step, someone could simply hold a photo of the target person in front of the camera and thereby cheat the check. The liveness check makes this trick considerably harder.

Liveness checks as a protective layer in digital identity verification

Facial recognition alone is not enough to prevent identity fraud. A system that only checks whether a face matches an ID photo can be tricked with a printed photo or a played-back video. Such attacks are known in technical jargon as spoofing attacks – that is, attempts to deceive the system with a fake input. The liveness check is the direct answer to this.

Requirements are particularly high in the financial sector. Banks and payment service providers are legally required in many countries to securely verify the identity of their customers – this is called Know Your Customer, or KYC for short. A passed liveness check is often mandatory in this process. If it is missing or too weak, the company risks fines and, in the worst case, enables money laundering through fake accounts.

Passive and active detection of real persons

There are two basic methods. In active verification, the person is asked to do something: turn their head, blink, or smile. The idea behind this is simple – a photo cannot imitate this. A video could, in principle, but a recorded video cannot adapt in real time to an unpredictable command.

Modern systems increasingly rely on passive verification. In this approach, the person doesn’t have to do anything at all. Instead, an AI model – that is, a system trained using examples – analyzes tiny features in the image: the texture of the skin, natural micro-movements, the way light reflects off a real surface. A printout or a screen does not display these features in the same way. Passive methods are more convenient for users but technically more demanding.

With the rise of so-called deepfakes – realistic-looking, AI-generated facial videos – the liveness check is becoming an arms-race task. Attacks are becoming more sophisticated, so detection models must be continuously retrained. For this reason, some newer systems combine several signals simultaneously: camera image, infrared light, and depth information from a 3D sensor, as is built into modern smartphones.

Liveness checks in apps, government agencies, and the news

In everyday life, one encounters the liveness check mainly during online account opening. Anyone opening an account with a direct bank in Germany usually goes through a VideoIdent or AutoIdent procedure. In this process, you film yourself and hold your ID up to the camera. The liveness check runs in the background, often without the user noticing a separate step.

Government bodies also use this procedure. The European eIDAS Regulation – a set of rules for digital identities in the EU – requires that identity verifications be tamper-proof for certain levels of trust. Liveness detection is recognized there as an acceptable means. Similar requirements apply in the USA for government agencies and financial regulators.

The term appears regularly in the tech press when it comes to security vulnerabilities. Researchers repeatedly publish findings that certain liveness checks have been tricked using low-cost deepfake videos. Such reports drive the development of new standards. At the same time, criticism from the privacy and data protection community is growing: facial images and biometric data – that is, bodily characteristics used as a digital key – are considered especially sensitive, and a poorly secured system can put exactly this data at risk.