CyberLinkprovides FaceMe® AI facial recognition technology,purpose-built as an anti-fraud line of defense for bank ATMs. The system supports Edge AI processing and can accurately detect face coverings such as masks or helmets,face coverings,and can integrate with access control systems to help financial institutions effectively blockscam syndicate mule behavior,safeguarding the public's financial assets.

In an Era of Labor Shortages, AI Facial Recognition Becomes a Key Tool for Real-Time Fraud Prevention at Financial Institutions!

In recent years, scam syndicate mules have often deliberately worn masks or helmets while withdrawing cash at ATMs to avoid identification, making it harder for police to track them down. To strengthen anti-fraud measures, Taiwan's National Police Agency has proposed to the Financial Supervisory Commission a"face covering alert"policy recommendation.

When a person approaches an ATM to withdraw cash,the AI facial recognitionsystem detects their facial condition in real time;if the face is covered and full facial features can't be identified,the system sounds an alert prompting the user to show their full face.
For mules deliberately covering their faces, this real-time prompt creates noticeable psychological pressure, achieving deterrence at the source.

For ordinary users, there's no need to worry.Simply removing a mask, sunglasses, or helmet temporarily during the transaction allows it to proceed smoothly, with no extra hassle.

Proven Results | Significant Drop in Mule Withdrawal Incidents 42.2%

According to pilot data from Chunghwa Post and the Tainan City Government (starting May 2025), after AI face-covering detection and real-time voice alerts were deployed at select locations, clear results were observed within five months:

• Mule withdrawal incidents dropped 42.2% year over year
• The number of police and prosecutor footage requests dropped correspondingly
• No added burden for users or extra enforcement cost

Currently, 6 financial institutions have joined the initiative, with over 500 anti-fraud ATMs now officially live across Taiwan.

The figures above are statistics from a government pilot program, intended only as a reference case for anti-fraud mechanism adoption, and do not correspond directly to any specific brand or product.This shows that real-time detection combined with on-site alerts can effectively lower the rate of fraudulent withdrawals.

Upgraded Anti-Fraud Mechanism | CyberLink's Face-Covering Detection Design

1 | First Line of Defense at the Entrance: Real-Time Edge AI Monitoring

Using Edge AI technology, detection and control are completed on a single device without sending video back to a remote server, resulting in faster response times.

• Automatic Facial Feature Scan:Accurately detects whether the eyes, nose, and mouth are fully visible
• Smart Access Control Integration:If a covered face is detected, the entry point will temporarily remain locked
• Multiple Voice Prompts or Light Indicators:Prompts such as "Please remove your mask" or "Please show your full face" as needed

2 | Privacy-Friendly Design: Detects Only Whether Features Are Covered, Not Identity

The system only determines whether facial features are fully visible, and doesnot match against personal identity data,balancing security with privacy protection.

Once a user shows their full face, they can proceed normally into the withdrawal area with no added hassle.

• Protects user privacy
• Effectively blocks fraud mules who deliberately cover their faces

3 | Multi-Layered Protection: Building a Complete ATM Anti-Fraud Process

• Layer 1 | Precise Entry ControlFacial Recognition Access Tablet:Entry remains locked for anyone not showing their full face, with voice prompts guiding them to comply.
• Layer 2 | 24/7 Real-Time MonitoringCompact HD Cameras:Continuously record and store footage that can be cross-referenced with withdrawal records to improve investigation efficiency.
• Layer 3 | Active On-Site DeterrenceSmart Alert Speaker: Immediately emits a high-decibel alert at suspicious individuals (mules) who deliberately cover their faces,breaking down criminal intent on the spot.
• Layer 4 | Real-Time Monitoring ResponseCentral Management Interface: Displays anomalies in real time, helping security or on-site staff quickly grasp the situation.
• Layer 5 | Event Traceability and AuditingIncident Investigation System:Integrates withdrawal records with video footage to provide a complete basis for post-incident tracing and evidence gathering.