Why did my bank flag my video selfie as fake?

It is a uniquely frustrating experience: you are a real, living person trying to open a bank account or verify your identity, and a machine tells you that you are not. The video selfie, designed to be a simple and secure step, has instead become a barrier. If your video selfie verification failed even though you are a real person, you are not alone. This issue stems from the complex challenge of "liveness detection," the technology designed to distinguish a live human from a photo, a mask, or a sophisticated digital deepfake. Understanding how these systems work, and why they sometimes fail, is the first step toward a more secure and less frustrating digital identity landscape.

"Even with modern systems, only about 85-90% of legitimate users pass a liveness verification check on their first attempt. While retries can increase this to over 98%, a significant number of real people are initially and incorrectly flagged as potential fakes."

When a video selfie verification failed for a real person

The core reason a video selfie verification failed real person scenario occurs is because of a mismatch between what the system expects and what it sees. These systems are tasked with performing Presentation Attack Detection (PAD), the technical term for identifying and rejecting attempts to "present" a fake likeness to the camera. These attacks can range from holding up a printed photo to using a hyper-realistic 3D mask or a digital deepfake video. To counter this, verification systems look for signs of life. However, not all methods are created equal, and many factors can lead to a false rejection of a genuine user.

Common reasons for failure can be grouped into two categories: environmental/user factors and system limitations.

• Poor Lighting: Heavy backlighting, shadows, or a room that is too dark can obscure the subtle facial features the algorithm needs to analyze.
• Camera Quality: An older, low-resolution webcam or a dirty smartphone lens can produce a blurry or grainy image that the system cannot process.
• Facial Obstructions: Hair covering the eyes, glare on glasses, or even a hat can cause the system to fail, as it cannot get a clear view of the entire face.
• Micro-movements: Many passive systems look for natural, subtle movements and color changes in the skin. If you are too still, or the video is choppy, these signals may be absent.
• System Demands: "Active" liveness systems that require you to blink, smile, or turn your head can be prone to failure if the instructions are unclear or if you don't perform the action in the exact way the machine expects.

These issues create a high False Rejection Rate (FRR), which is a major point of friction for customers and a significant cost for banks and financial institutions that lose good customers during onboarding.

Feature	Active Liveness Detection	Standard Passive Liveness	rPPG-Based Passive Liveness
User Action	Required (e.g., blink, turn head, smile)	None (just a short video selfie)	None (just a short video selfie)
Primary Analysis	Challenge-response, motion tracking	Texture analysis, 3D mapping, eye movement	Photoplethysmography (blood flow)
User Friction	High	Low	Low
Vulnerability	Can be spoofed by sophisticated replays	Vulnerable to high-res masks & deepfakes	Extremely high resistance to all presentation attacks
Key Weakness	High user drop-off, frustrating UX	Can misinterpret non-standard faces or lighting	Requires a capable camera and sufficient light

Industry applications of liveness detection

The financial and security sectors have a critical need to ensure the person creating an account or accessing a system is who they claim to be and is physically present at the moment of verification. The stakes are incredibly high, involving the prevention of account takeover fraud, money laundering, and synthetic identity fraud.

### preventing account opening fraud

For banks, neobanks, and fintech platforms, remote account opening is the primary customer acquisition channel. It is also a massive target for fraud. Liveness detection serves as a digital doorman, ensuring that a fraudster can't open an account using a stolen ID and a simple photograph.

### securing high-value transactions

When a user requests a large wire transfer or changes their login credentials, many platforms now require a quick video selfie to re-verify their identity. This step-up authentication prevents account takeovers, where a criminal gains access to an account and attempts to drain the funds.

### Compliance with KYC/AML Regulations

Know Your Customer (KYC) and Anti-Money Laundering (AML) regulations require financial institutions to verify the identities of their clients. Robust liveness detection provides a much stronger audit trail than document-based verification alone, proving that a live person was associated with the identity documents provided.

Current research and evidence

The limitations of traditional liveness detection have spurred significant academic and commercial research into more robust methods. The most promising of these is remote photoplethysmography (rPPG), a technique that analyzes video frames to detect the microscopic color changes in human skin caused by the flow of blood. This is the same core technology used in hospital pulse oximeters, adapted for standard cameras.

Early pioneers in the field, such as Li et al. (2018), proposed methods to extract pulse signals from video to differentiate between live faces and photo attacks. Subsequent research, like the work of Liu et al., has advanced the field by applying deep learning models, such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), to more accurately estimate the rPPG signal and reject sophisticated spoofing attempts. Their work demonstrated how these models could learn to distinguish the characteristic patterns of a real human pulse from digital noise or the static texture of a mask. This approach moves beyond analyzing what a person looks like and instead verifies a fundamental physiological sign of life.

The future of identity verification

The future of digital identity is a move away from disruptive, high-friction challenges and toward silent, passive, and highly accurate verification. Systems based on rPPG blood flow detection represent the leading edge of this shift. By looking for a biological signal that cannot be faked by current generative AI or physical artifacts, these systems can dramatically reduce both False Rejections (letting real people in) and False Acceptances (keeping fraudsters out). For the user who was wrongly flagged as fake, this technology means a smoother, less frustrating experience. For the bank, it means stronger security, higher customer conversion, and greater trust in their digital platform.

Frequently asked questions

What is the most common reason my video selfie was rejected? The most common reasons are environmental. Poor lighting (too dark, or bright light behind you), having part of your face covered by hair or glasses glare, or moving the camera too much can all cause a failure. Try to be in a well-lit room with light facing you, keep your face clear, and hold your phone steady.

Is it possible for a deepfake to pass a liveness test? It is possible for a sophisticated deepfake or 3D mask to fool simpler liveness detection systems, particularly those that rely only on facial movement or texture. This is why the industry is moving toward more advanced methods that analyze physiological signs, like the presence of a human pulse.

How can I make sure my next video selfie verification is successful?

• Find a room with good, even lighting. Face the light source.
• Make sure your camera lens is clean.
• Hold your phone or position your webcam at eye level and keep it steady.
• Remove any hats, sunglasses, or anything that obscures your face. If you wear glasses, try to angle your head to avoid glare.
• Follow the on-screen instructions carefully and precisely.

A new generation of identity verification technology is addressing the core issue of liveness detection accuracy. Circadify is at the forefront of this space, developing solutions that move beyond fallible visual checks to analyze intrinsic physiological signals. By confirming the presence of a human pulse, this approach offers a fundamentally more secure and reliable way to prevent fraud while ensuring genuine customers are not turned away. If you are a business looking to improve the accuracy of your identity verification process, learn more about our advanced fraud detection solutions at circadify.com/solutions/fraud-detection.