Presentation Attack Detection (PAD)

Presentation attack detection is what stops a biometric check from being fooled by something held up to the camera or sensor. A presentation attack is any attempt to defeat biometric verification with a spoof: a printed photo, a video replay, a silicone mask, or an injected deepfake. PAD is the countermeasure, and the international standard for it is ISO/IEC 30107.

The discipline splits roughly into confirming a live human is present (often called liveness detection) and detecting the specific artifacts of known spoofs. Methods range from prompting an action the user performs in real time, to passive analysis of texture, depth, and signal cues that a fake lacks. Injection attacks, where a deepfake is fed straight into the data stream rather than shown to a camera, are a newer and harder class.

In CIAM, PAD is the defense behind the biometric step of identity proofing. As generative AI makes convincing fakes cheap, the strength of presentation and injection attack detection has moved from a footnote to a core evaluation question for any platform that verifies a face or document.

Sources

• ISO/IEC 30107-1, Biometric presentation attack detection: https://www.iso.org/standard/83828.html