Key takeaways
- Presentation attacks include prints, screen replays and masks.
- Passive liveness is low-friction; active liveness adds assurance.
- Injection attacks bypass the camera and need capture-integrity controls too.
- Look for evaluation against ISO/IEC 30107-3.
Why matching isn't enough
A fraudster with a victim's photo can hold it up to the camera. A pure match will happily say "same person". Liveness — formally presentation attack detection — catches that before the match counts.
Types of attack
- Print attacks: a photo on paper, sometimes with the eyes cut out.
- Replay attacks: a video played on a phone or monitor.
- Mask attacks: paper, silicone or 3D-printed masks.
- Injection attacks: synthetic or deepfake video fed directly into the app or a virtual camera, never touching a lens.
Passive vs active
Passive liveness analyses a single frame or short clip for texture, depth and reflection cues — the user just takes a selfie. Active liveness asks for an action like turning the head. Many teams run passive by default and step up to active for risky sessions.
Standards and testing
ISO/IEC 30107-3 defines how presentation attack detection is tested and reported. When evaluating vendors, ask what attack types were tested, by whom, and at what error rates.
Best practices
- Run liveness and matching on the same capture.
- Combine with device signals to resist injection attacks.
- Give clear capture guidance — lighting and framing matter.
- Route borderline scores to manual review rather than auto-rejecting.
Try it with the SIMSIM liveness API.
