Anti-Fraud Tool Goes Live! AI Facial Recognition Blocks Fraudulent ATM Withdrawals
Scams are constantly evolving. Fraud rings often lure victims into transferring money under the guise of investments, impersonated government agencies, or romance scams, then dispatch money mules to withdraw the stolen funds at ATMs.
Criminal Investigation Bureaureleased the top ten ATM fraud withdrawal hotspots nationwide for the first half of 2025, with most locations clustered around highway interchanges, MRT stations, or university campuses, open areas with easy access. Money mules typically wear masks throughout the withdrawal process to conceal their identity and avoid recognition. According to statistics, fraud losses totaled NT$8.878 billion in 2023 alone.
To address the growing fraud problem, Taiwan's Financial Supervisory Commission has recommended that banks deploy "anti-fraud" devices at ATMs, requiring users to show their full face, without masks or helmets, when withdrawing cash or making transfers. Many banks have already responded by rolling out the equipment. When a user hasn't removed their mask, the system immediately issues a voice and visual alert, effectively deterring money mules and adding an extra layer of protection for the public's finances.
AI Makes Fraud Prevention More Effective, Upgrading ATM Security!
ATM self-service areas typically have no access control, making them a prime target for fraud rings. CyberLink's anti-fraud solution, the Facial Recognition All-in-One Unit, is installed around ATMs and uses AI edge computing to automatically detect whether the user's full face is visible. The eyes, nose, and mouth must all be clearly visible before the door will open.
It also providesvoice, light, and on-screenprompts to quickly guide the public into compliance.
The entire process runs automatically, with no need for additional staff to be stationed at the ATM, significantly improving the security and fraud-prevention efficiency of self-service ATM areas!
Full-Face Recognition, Zero Privacy Leakage
As long as the full face is clearly visible, the system can make a quick determination without comparing it against any identity database.This design avoids excessive collection of personal data, protecting privacy while effectively blocking money mules who try to disguise themselves with masks, sunglasses, or helmets.
AI Facial Recognition Effectively Blocks Disguises
When the system detects that a user is wearing sunglasses, a mask, a helmet, or has hair covering their face, it automatically denies entry and issues a matching voice alert. If a mask is the cause, it plays "Please remove your mask"; for other obstructions, it prompts "Please show your full face," ensuring the withdrawal process stays secure and clear.
GIS YouMe 500 Product Highlights
- Designed and Manufactured in Taiwan, Quality Assured
- Powered by a MediaTek quad-core processor for strong, stable performance
- 3D liveness detection + CyberLink FaceMe algorithm, 98.95% recognition accuracy
- Under 1-second fast pass-through, supports a database of up to 50,000 people
- Versatile access control interfaces, compatible with various magnetic locks / gates
- LED indicators + voice prompts for an intuitive user experience