An Ninh Số deploys AI cameras for facial recognition & weapon detection for banks

A commercial joint stock bank in Ho Chi Minh City with 45 branches and 120 transaction offices faces a dual security challenge: it must comply with the SBV’s KYC/AML regulations while minimizing robbery/vandalism risks at ATMs. An Ninh Số — a member brand of Việt Đức Trí Group — deployed an integrated AI camera system featuring facial recognition, automatic weapon detection, and 24/7 ATM monitoring, completed in 14 weeks with over 1,800 devices installed.

Customer: Joint Stock Commercial Bank (name undisclosed per NDA) | Industry: Finance - Banking | Scale: 45 branches, 120 transaction offices, 180 ATMs | Location: Ho Chi Minh City and 12 southern provinces

Customer Challenges

• Failure to detect suspicious individuals in a timely manner: The old camera system only recorded passively, with no facial recognition or weapon detection capabilities. On average, 3-5 suspicious incidents occurred at transaction counters each month, but 80% were not detected until reviewing recorded footage.
• Ineffective manual ATM monitoring: 180 ATMs were guarded by rotating security personnel, costing an estimated 2.3 billion VND/year. However, 40% of ATM break-in attempts occurred outside guard duty hours, causing average losses of 120-180 million VND per incident.
• Slow and inaccurate KYC/AML compliance: The process for verifying customers transacting over 300 million VND took 15-20 minutes per instance, requiring staff to manually search databases. The identity error rate was approximately 5-7%.

Deployed Digital Security Solution

An Ninh Số deploys a three-layer AI camera system: layer 1 — facial recognition at transaction counters (Hikvision DeepinView), layer 2 — automatic weapon detection in main lobbies (Edge AI Dahua WizMind), layer 3 — 24/7 ATM surveillance with real-time alerts. All data is synchronized to the central VMS via a virtual private network infrastructure.

1. Survey & Design (2 weeks): Survey 45 branches, identify 280 priority installation locations (transaction counters, main lobbies, ATMs). Build network topology diagrams and calculate bandwidth (minimum 100 Mbps/branch).
2. Hardware Installation (8 weeks): Deploy 1,840 cameras: 720 Hikvision DS-2CD2T87G2-LSU/SL cameras (4K, AI edge) for transaction counters, 560 Dahua N85CZ6 cameras (5MP, weapon detection) for main lobbies, 560 Hikvision DS-2CD2T47G2-L cameras (4MP, infrared) for ATMs. Install an additional 180 32-channel NVRs (Hikvision DS-9632NXI-I8/S).
3. Software Integration (3 weeks): Configure AI engine on NVR: facial recognition (Face Recognition 2.0, 98.7% accuracy), weapon detection (Weapon Detection — fine-tuned YOLOv8 model, 85% threshold), ATM surveillance (Intrusion Detection + Loitering Detection). Synchronize data to central VMS via dedicated VLAN.
4. Testing & Acceptance (1 week): Run 500 test scenarios (fake faces, fake weapons, ATM intrusion). Achieve 96.2% correct detection rate, false alarms under 3%.

Achieved Results

• 65% — incident response time reduced (from 45 minutes down to 15-18 minutes)
• 96.2% — correct suspect detection rate
• 40% — ATM security costs reduced (saving 920 million VND/year)
• 14 months — return on investment (ROI) period

After 6 months of operation, the system detected 12 ATM tampering incidents (timely prevention of 10 incidents), identified 47 wanted individuals at transaction counters, and reduced KYC verification time by 85% (from 15-20 minutes down to 2-3 minutes per transaction).

“Previously, every time there was an incident at an ATM, security personnel took 30-45 minutes to arrive. Now, An Ninh Số’s AI system automatically sends alerts with suspect photos to the shift supervisor’s phone in just 2-3 seconds. We prevented 2 ATM tampering cases in the first month. For the bank, that’s a saving of at least 300 million VND.”

— Mr. Nguyen Van Hung, Director of Operations & Security Division, Joint Stock Commercial Bank (name withheld per NDA)

Lessons learned for financial industry businesses

This project shows that AI cameras are not just passive recording tools but proactive security systems capable of real-time analysis and deep integration into operational processes. Financial businesses need to note three factors when deploying: (1) choose cameras with AI edge to reduce bandwidth load, (2) invest in a private network infrastructure to ensure data security, (3) train the AI engine with real-world data (do not use generic models) to achieve over 95% accuracy.

To learn more about AI camera solutions for banks, you can refer to the article on AI Vision for automatic payment — a similar technology being applied in the retail sector. Or check out the case study ERP integrated with POS to see how Việt Đức Trí Group digitally transforms businesses comprehensively.

Frequently Asked Questions

Does facial recognition AI camera violate privacy?

The system complies with Decree 13/2023/ND-CP on personal data protection. Facial data is encrypted with AES-256 and stored on internal servers (not public cloud). Banks are responsible for notifying and obtaining customer consent when installing cameras at transaction counters. An Ninh Số provides an automatic face-blurring option for customers who do not consent.

What is the investment cost for AI cameras at a bank branch?

The average cost for a branch with 5-7 transaction counters and 2-3 ATMs ranges from 180-250 million VND (including AI cameras, NVR, network cables, installation). For a scale of 45 branches like the project above, the total investment is about 8-11 billion VND, with an ROI of 14 months thanks to savings on security guard costs and loss reduction.

Can the system detect weapons hidden in bags?

AI cameras use weapon detection technology based on shape and metallic materials (combined with infrared sensors). The correct detection rate reaches 96.2% for guns, knives, and scissors longer than 15 cm. However, weapons hidden deep inside thick bags or suitcases may not be detected — in this case, the system will record abnormal behavior (loitering, strange gestures) to issue alerts.

Does An Ninh Số provide system maintenance after deployment?

Yes. An Ninh Số offers a 12-month maintenance package (free for the first year) including: periodic checks twice a month, AI engine firmware updates, and replacement of faulty components within 24 hours. After the first year, the maintenance fee is approximately 8-12% of the contract value per year.

This project demonstrates An Ninh Số’s capability to deploy large-scale AI cameras — a member brand of Việt Đức Trí Group. If your bank/credit fund is looking for a proactive security solution, please contact the hotline 0935 295 337 immediately for a detailed deployment roadmap consultation.