Presentation Attack Detection (PAD) is the technical term for anti-spoofing in biometrics — methods that detect when someone presents fake audio (a recording, synthetic clip, or deepfake) to a voice system instead of speaking genuinely. PAD is defined by the international standard ISO/IEC 30107 and is what lets voice biometrics resist spoofing.
If you evaluate voice biometrics, you will encounter the term “PAD.” It is the formal name for the safeguards that stop voice spoofing. Here is what PAD means and how it protects against deepfakes.
What is a presentation attack?
A presentation attack is any attempt to defeat a biometric system by presenting a fake or artificial input — in voice biometrics, that means a recording, a text-to-speech clip, a cloned voice, or converted audio, rather than the genuine live user speaking.
How PAD works
- Artifact analysis. PAD models detect the statistical signatures left by recorded or synthetic audio.
- Liveness checks. Confirming a live speaker is present in real time.
- Challenge-response. Randomized prompts that a pre-recorded or pre-generated sample cannot satisfy.
PAD standards: ISO/IEC 30107
PAD is governed by the international standard ISO/IEC 30107, which defines presentation attacks and how to measure a system’s ability to detect them. Referencing recognized standards helps organizations compare solutions on a consistent basis.
PAD and deepfakes
Deepfakes are among the most advanced presentation attacks, so strong PAD is essential. VoiceVantage’s PAD capability was demonstrated when VoiceCheck blocked all five leading deepfake applications across 533 independent University tests.
Frequently Asked Questions
Is PAD the same as anti-spoofing?
Yes. “Presentation attack detection” is the formal, standards-based term for anti-spoofing in biometrics.
What is ISO/IEC 30107?
It is the international standard that defines biometric presentation attacks and how to evaluate detection of them.
Why does PAD matter for voice biometrics?
Without PAD, a system could be fooled by recordings or deepfakes. With it, those attacks are detected and blocked.
Evaluating voice biometrics? Ask VoiceVantage about our PAD results across 533 independent tests.