Liveness detection
Liveness detection determines whether a biometric sample comes from a live person present at the sensor, rather than a photograph, printed mask, screen replay or deepfake. It is the defence against presentation attacks, and international standard ISO/IEC 30107-3 defines how such defences are tested and reported.
What is a presentation attack?
A presentation attack is any attempt to defeat a biometric system by presenting a fake sample to the sensor. In face recognition the common forms are a printed photograph, a face displayed on a phone screen, a video replay, a paper or silicone mask, and increasingly a synthetically generated face injected into the camera feed.
- Print attack: a photograph of the target held up to the camera.
- Replay attack: a video of the target played on a screen.
- Mask attack: a physical mask, from paper cutouts to silicone.
- Injection attack: synthetic or captured video fed directly into the video pipeline, bypassing the camera entirely.
Active versus passive liveness detection
| Active | Passive | |
|---|---|---|
| What the user does | Blink, turn, smile on request | Nothing |
| Time to complete | Seconds | Instant |
| Works on existing CCTV | No | Yes |
| Resists replay of a recorded challenge | Weaker | Not applicable |
Active liveness asks the user to perform an action and checks that it happened. Passive liveness analyses a single capture for the texture, depth and reflection cues that separate a real face from a reproduction. Passive is the only option where recognition happens without cooperation, such as a camera at a gate.
What does ISO/IEC 30107-3 measure?
ISO/IEC 30107-3 is the international standard for testing presentation attack detection. It defines the attack instruments a system must be tested against and the metrics used to report results, so that two vendors' claims can be compared rather than each defining success for itself.
Where the check runs matters as much as how well it scores. Liveness analysis depends on fine texture and reflection detail, so a system that evaluates a heavily compressed frame pulled from a recording has already discarded most of what the check reads. Running it on the capture before compression, at the edge, is what keeps a laboratory result relevant to the installed system.
A vendor claiming to be ISO/IEC 30107-3 compliant is describing a test methodology, not a pass mark. Ask which attack instruments were used, at which level, and who performed the testing.
Frequently asked questions
- What is liveness detection?
- Liveness detection determines whether a biometric sample comes from a live person at the sensor rather than a photograph, mask, screen replay or deepfake. It is the defence against presentation attacks, and ISO/IEC 30107-3 defines how those defences are tested and how results are reported.
- What is the difference between active and passive liveness?
- Active liveness asks the user to blink, turn or smile and verifies the action occurred. Passive liveness analyses a single capture for texture, depth and reflection cues without asking anything of the user. Passive is the only workable option where subjects are not cooperating, such as a camera at a gate.
- Can liveness detection stop deepfakes?
- Presentation attack detection addresses fakes shown to a camera, including a deepfake played on a screen. It does not by itself address injection attacks, where synthetic video is fed directly into the pipeline and never passes a lens. Injection defence requires securing the capture path, not only analysing the image.
- Does liveness detection work on existing CCTV cameras?
- Passive liveness can operate on standard camera streams, subject to resolution and lighting. Active liveness cannot, because it requires an interaction loop with the subject and a screen to prompt them, which a fixed surveillance camera does not provide.
- Is ISO/IEC 30107-3 a certification?
- No. It is a testing and reporting standard that defines attack instruments and metrics so results can be compared between vendors. A claim of compliance describes the methodology used, not a pass mark awarded, so ask which attack instruments were tested, at what level, and by whom.
- Does liveness detection slow down a gate or a door?
- Passive liveness adds one analysis pass to a frame the system already captured, so throughput is governed by the recognition pipeline rather than by the check itself. Active liveness adds seconds per person because it waits for a human response, which is why it is rare on high-traffic entrances.