On this page
What is a presentation attack?
A presentation attack is an attempt to fool a biometric sensor by showing it a fake sample. The standard calls that fake a presentation attack instrument, or PAI, and the term covers anything held up to the camera in place of a real face.
The cost range is the interesting part. The cheapest attack needs a public photo and a color printer. The expensive ones need a mold of somebody's face. A system that stops the first kind and not the second is still useful, as long as you know which one you bought.
- A printed photo of a face
- A video of a face on a screen
- A paper mask or a 3D mask
- A deepfake face on a display
What does the ISO/IEC 30107-3 standard define?
ISO/IEC 30107-3 is the international standard for how presentation attack detection is tested and reported. It does not tell a vendor how to build a detector. It defines the experiment: which artifact types are used, how many attempts are made, and how the result is expressed.
That is why an accredited laboratory result carries weight and a self-assessment does not. The companion standard ISO/IEC 30107-1 holds the taxonomy the test relies on, including the PAI vocabulary above.
What do the PAD levels mean?
Level 1 and Level 2 are iBeta program tiers. ISO/IEC 30107-3 does not define them, which surprises people who see the two cited together in one sentence.
Each tier caps the cost and sophistication of the artifacts a laboratory may build. A higher tier means the testers were allowed to spend more per attack, not that the standard changed.
Level 1 uses low-cost artifacts
Printed photographs, images and video replayed on a screen, and simple paper masks. These are the attacks any motivated person can mount today with almost no budget.
Level 2 uses custom artifacts
Made-to-order items such as a fitted 3D mask, produced by the laboratory for the specific test subject. This is the tier where the cost of an attack starts to deter one.
How does PAD relate to liveness detection?
Liveness detection is the everyday name and presentation attack detection is the standards name. When a datasheet uses the formal term, it is usually because there is a laboratory result behind it.
The liveness detection page covers the same ground from the practical side, including how passive and active checks differ in use.
How does SenseCrypt use presentation attack detection?
SenseCrypt signs a user in by face, so PAD sits on the critical path of every sign-in. Capture, liveness, and matching all run on the user's own device, and a capture that fails the check never produces a face token.
The SenseCrypt liveness holds iBeta ISO 30107-3 certification covering Level 1 and Level 2. Face recognition is a separate question. It entered the NIST evaluation in 2021, which was then FRVT and later split into FRTE and FATE, and is maintained through our latest submissions. We call that evaluated, not certified.
PAD gates the token
A failed check ends the ceremony on the device. No face token is minted, so nothing reaches the identity provider to be judged after the fact.
Tested at both tiers
The iBeta result covers Level 1 artifacts such as a printed photo or a screen replay, and Level 2 artifacts such as a made-to-order 3D mask.
Nothing stored to attack
The server stores no face image and no face template, so a successful attack finds no gallery to poison and no template to steal.
Frequently asked questions
What is presentation attack detection?
Presentation attack detection (PAD) is a method that finds and blocks fake biometric samples at a sensor. The ISO/IEC 30107-3 standard defines PAD.
What is the difference between PAD and liveness detection?
The two terms share one goal. Presentation attack detection is the formal term in ISO/IEC 30107-3. Liveness detection is the common name.
What do PAD Level 1 and Level 2 mean?
Level 1 covers simpler attacks, such as a printed photo. Level 2 covers stronger attacks, such as a custom 3D mask.
Does SenseCrypt hold a PAD certification?
Yes. The SenseCrypt liveness detection holds iBeta ISO 30107-3 certification. It covers Level 1 and Level 2 presentation-attack detection.
Related