Skip to content

Facial Recognition Systems in Singapore: How They Work, Where They Help and PDPA Rules

Updated 19 Nov 20266 min readBy CCTVSG.NET, Singapore

In short

How a facial recognition system works, where Singapore homes and businesses use it (door access, alerts, visitors), its accuracy, and the PDPC's biometric rules.

A facial recognition system compares a face captured by a camera or door terminal with a list of faces you have enrolled, then acts on a match: it opens a door, logs who came in, or sends you an alert. In Singapore it is used for office door access, staff and visitor lists, and alerts when someone on a watch list appears. Businesses may use it for security under the PDPA if they tell people, protect the face data and delete it when it is no longer needed, as the PDPC’s 2022 guide on biometric data sets out. Our face recognition CCTV package, AI+, starts from S$980.

Key Takeaways

  • Face recognition does not store a photo for matching; it turns each face into a template, a set of numbers that only that system can compare.
  • Use a door terminal for access and a face recognition camera for alerts at a distance; they solve different problems.
  • Matching is probabilistic: set the threshold to match the risk, and add a card or PIN where a wrong match would be costly.
  • Under the PDPA, employers can rely on the employment exception for staff access, but must still tell employees how their face data is used.
  • Put up notices, keep templates encrypted, and delete face data once the person leaves or the purpose ends.

How does a facial recognition system work?

A facial recognition system is software, running in a camera, a recorder or a door terminal, that identifies or verifies people by comparing their faces with enrolled faces. It works in four steps:

  1. Enrolment. You add each person once, with a clear front-facing photo taken at the terminal or uploaded.
  2. Template. The system turns the photo into a biometric template. The PDPC describes templates as data produced by technical processing of the face image, which application systems then use for matching.
  3. Capture. When someone walks up to a terminal or past a camera, the system finds the face in the image and makes a new template.
  4. Match and act. It compares the new template with the enrolled ones and, above a set similarity score, opens the door, records the entry or sends an alert.

There are two kinds of matching. Verification (one-to-one) checks that you are the person you claim to be, for example a face plus an access card. Identification (one-to-many) searches the whole list to find who you are, which is what a hands-free door or a watch-list camera does.

Where is facial recognition used in Singapore?

UseDeviceTypical site
Hands-free door accessFace recognition terminal at the doorOffices, clinics, server rooms
Entry records and attendanceFace terminal with time-stamped logsOffices, factories, worksites
Alerts for listed people (blacklist)Face recognition camera and recorderShops, showrooms, building lobbies
Knowing when staff or family arrive (whitelist)Face recognition cameraHomes, small offices
Finding a person in recordingsRecorder with face searchAny site that investigates incidents

Door terminals are covered in our guide to card, fingerprint and face door access. For offices planning access for the whole team, see the office access control checklist.

Face recognition camera or face recognition terminal?

FeatureFace recognition terminalFace recognition camera
DistanceClose: the person looks at the screenA few metres: people walk past
Main jobOpen a door for enrolled peopleAlert on listed faces, record who passed
People's cooperationYes, they stand in front of itNo, faces are captured as people move
AccuracyHigher: steady face, good lightLower: angle, speed, caps and masks matter
Our packageFace recognition access from S$680 per doorAI+ from S$980

How accurate is facial recognition?

No face system is perfect, because matching works on probability. The PDPC’s guide names two errors: a false negative, where an enrolled person is not recognised because the matching threshold is set too high, and a false positive, where the system wrongly identifies someone else as the enrolled person because the threshold is set too low. It advises setting the threshold according to the impact of a wrong match, for example higher for entry into a data centre, and adding a second factor such as an access card where you need more assurance.

These things affect accuracy the most:

  • Angle. Mount face cameras at a height and angle that see faces from the front, not the top of heads.
  • Light. Strong light behind people at a glass door turns faces into silhouettes; choose a spot or camera that handles backlight.
  • Enrolment photos. One clear, front-facing photo per person beats several poor ones.
  • Caps, masks and movement. A camera covering a busy walkway will miss some faces; a terminal at the door will not.

The guide also warns about spoofing, such as holding up a photo or a 3D model of an enrolled person’s face. It recommends anti-spoofing measures such as liveness detection, and placing face access points near a manned security post where possible.

What are the PDPA rules for facial recognition in Singapore?

A face image or template linked to a person is personal data, so the Personal Data Protection Act applies when a business collects it. The PDPC’s Guide on the Responsible Use of Biometric Data in Security Applications (May 2022) sets out how this works for security uses:

  • Staff: employers may rely on the employment exception for biometric access control of staff entry, but should still tell employees how their biometric data will be used, for example in the employee handbook or contract.
  • Visitors and members of the public: organisations should obtain consent for collecting biometric data unless an exception applies. For cameras in places open to the public, the guide gives the publicly available data exception, with notices as good practice. For areas not open to the public, an organisation may rely on the legitimate interests exception if it meets that exception’s requirements, including an assessment.
  • Protect the templates: process face photos into templates as soon as practicable, use only the templates for matching, and encrypt them.
  • Delete when done: biometric data should not be kept once its purpose is fulfilled and there is no business or legal need to keep it, for example after an employee leaves.
  • Access requests: a person may ask for the face images you hold of them. The guide says you need not hand over the template itself.

The guide is not legally binding, but it shows how the PDPC applies the PDPA to biometric systems. The guide does not cover personal or household use, so a homeowner’s own face camera is outside it. Our PDPA CCTV guide covers notices and retention for all CCTV. This is general information, not legal advice.

How to set up facial recognition at your premises

  1. Write down the purpose: open a door, alert on a blacklist, or record who enters. It decides the device.
  2. Pick the spot: a doorway where people face the camera, with even light and a short walk.
  3. Enrol people with clear photos, and record when and why each person was added.
  4. Set the threshold and alerts, test with your own staff for a week, and adjust.
  5. Write the data rules: notice wording, who can view the list, and when to delete faces.

Our face recognition packages

Our AI CCTV package AI+ is a Hikvision face recognition CCTV system that checks faces against your whitelist or blacklist, captures and compares faces, detects line crossing and intrusion, and sends app and email alerts. It starts from S$980. For hands-free entry, our Hikvision face recognition door access starts from S$680 for a single door on the door access page. Since 2011, CCTVSG.NET has installed CCTV and door access across Singapore. Request a quote or visit our showroom to try face recognition before you decide.

Sources

Frequently asked questions

Is facial recognition legal in Singapore?
Yes, for legitimate purposes such as security, if the organisation follows the PDPA: get consent unless an exception applies, tell people with notices or the staff handbook, protect the face data and delete it when the purpose ends. The PDPC's 2022 biometric guide explains how.
Do I need my employees' consent for face recognition door access?
The PDPC's biometric guide says employers may rely on the employment exception for biometric access control of staff entry, but should still tell employees how their biometric data will be used, for example in the employee handbook.
Does face recognition work with masks or caps?
Accuracy drops when part of the face is covered or the face is turned away. A door terminal, where people stop and face the screen, copes better than a camera covering a busy walkway.
How much does a facial recognition system cost in Singapore?
Our Hikvision face recognition door access starts from S$680 for a single door, and our AI+ face recognition CCTV package starts from S$980. The final price depends on the number of doors, cameras and the site.

Want this done for you?

We supply and install CCTV, door access, video intercom and Wi-Fi systems across Singapore, from Hikvision, Uniview, Vstarcam, TP-Link, Ruijie and EZVIZ. Send a photo or floor plan and we will recommend a setup and a fixed price.

CCTV CAMERA PACKAGES PROMOTION

Package 1 - Hikvision Dome or bullet CCTV Camera

Package 1 - Hikvision Dome or bullet CCTV Camera - 1 camera - Option A
OR
Package 1 - Hikvision Dome or bullet CCTV Camera - 1 camera - Option B

Package 4 - Vstarcam Wireless IP Camera

Package 4 - Vstarcam Wireless IP Camera - 1 camera - Option A
OR
Package 4 - Vstarcam Wireless IP Camera - 1 camera - Option B

Need CCTV installed in Singapore?

Tell us about your home, shop or office and get a fixed-price quote, usually within a few hours. 15+ years of installations, licensed by the Singapore Police Force.

WhatsAppGet quoteCall