How to stop 99.5% of selfie video tampering with AI

Selfie videos used to be a foolproof way of ensuring that your clients are who they say they are, for example when opening a new bank account. Clients are prompted to record a short video of their face while following a set of instructions such as looking from left to right. The videos are then used to verify the person behind the device.

However, with every leap forward in verification techniques comes a corresponding advance in fraud. By using fake videos, fraudsters can get around this step and gain access to your clients’ accounts. This poses a significant threat to businesses that rely on online identity verification for security, compliance, and client trust. It exposes them to potentially significant financial losses and reputational damage.

To address this challenge, Fourthline has built a proprietary system to detect selfie video tampering to enable financial service providers to ensure the integrity of their identity verification processes and stay ahead of fraudsters.

Understanding selfie video tampering

Selfie video tampering is where an individual submits a video for identity verification that does not meet the real-time requirements of the verification process. There are two main types of selfie video tampering:

Pre-recorded videos

Fraudsters use previously generated photographic or video material to falsify an identity verification in real time. They use various media from paper to TV, monitors, smartphones, and tablets to impersonate a client and gain unauthorized access to their funds.

Counterfeit videos

Fraudsters use advanced facial morphing and deepfake technologies to alter the facial features or voice in a video. This can be achieved by overlaying another person's facial features onto their own or by digitally altering video materials to mimic a real-time recording.

How Fourthline ensures liveness during selfie video verification

Our system uses machine-learning algorithms to sift through selfie video submissions and detect pre-recorded videos or counterfeit content, such as paper masks, inanimate objects, and a variety of other anomalies, to ensure that only authentic videos pass the verification process. The result is that 99.5% of fraudulent videos are flagged and blocked, while 97.5% of authentic videos pass through seamlessly.

Key features of our video verification technology

Powerful machine-learning algorithms: Our system uses advanced machine learning, which is capable of assessing a wide range of videographic and biometric cues.

Real-time processing: Videos are processed within seconds, allowing for fast and efficient identity verification without compromising on security or compliance.

Quality data, quality output: To achieve unparalleled accuracy in detecting fraudulent verification attempts, our models are trained on a high-quality and representative dataset, which has been meticulously assembled by our biometrics and annotation experts.

Customizable risk tolerance: Our system offers customizable risk thresholds, which you can adjust in accordance with your risk management strategy.

Modularity and flexibility: Our system can be combined with other checks to create a watertight fraud defence tailored to your specific use cases.

Seamless integration: Our system provides a frictionless user experience for both our business partners and your clients.