Active versus passive liveness
Active liveness asks the person to do something – blink, turn the head, smile, follow a dot on screen – and checks that the motion is natural. It is easy to explain to workers and hard to beat with a still photo, but it adds two to four seconds per punch and can fail for people wearing thick spectacles or standing in harsh sunlight.
Passive liveness analyses a single frame or a short burst without asking for any action. It looks at skin texture, reflections, depth cues and moiré patterns that appear when a camera photographs another screen. It is faster and less annoying at a busy gate, but relies more on the phone camera quality. Most attendance apps now combine both: passive by default, with an active challenge when confidence is low.
The spoof types it has to catch
The attacks an Indian employer actually sees are low-tech. A helper hands his phone with a WhatsApp DP of an absent colleague to a friend; a supervisor holds up a laminated ID-card photo; a guard replays a selfie video. High-end 3D masks are rare outside films. A good liveness layer therefore concentrates on print attacks, screen-replay attacks and cut-out masks, and flags them rather than silently rejecting.
Liveness does not verify identity. It only answers 'is this a live human'. Identity is the separate face recognition match against the encrypted descriptor stored at enrolment. Both must pass, and both should be logged, so that a rejected punch shows why it failed.
- Print attack: paper photo or ID-card photo held to the camera
- Replay attack: photo or video shown on another phone or tablet
- Mask or cut-out: a printed face with eye holes, worn by someone else
- Twin or look-alike: a genuine live face that is not the enrolled person – caught by the match step, not by liveness
Why it matters for selfie attendance and payroll
Without liveness, a selfie attendance system is just a photo log that someone has to inspect by eye at month end. With it, the punch is trustworthy at the moment of capture, which is what makes attendance-linked payroll and client billing defensible. It is also the control that closes the buddy punching loophole that fingerprint cards and PINs never solved.
Under the DPDP Act, face data is personal data; the employer should tell employees that a liveness check and face match are performed, what is stored (descriptors, not raw images) and for how long. Put this in the attendance policy rather than relying on a verbal briefing.
At 06:58 a housekeeping supervisor in Noida tries to mark in three absent staff by pointing his phone at their ID photos. The passive check flags flat texture on all three attempts and asks for a blink; nothing moves. All three punches are refused and logged as 'liveness failed' with a timestamp. The live monitor shows the three as absent at 07:15, and the operations head calls the site instead of discovering the gap on the client's bill.
Attend Mitra's face-recognition and selfie attendance include a liveness check before the face match. The app stores encrypted face descriptors rather than photos, records why a punch was rejected, and pairs the result with the GPS stamp and geofence rule for the site, so a supervisor can see failed attempts on the live monitor.