Voice Biometrics¶
Definition¶
Voice biometrics verifies identity based on unique vocal characteristics — pitch, tone, cadence, and vocal tract shape. In eKYC, voice can serve as an additional biometric factor during Video KYC or as a standalone verification method.
Voice in eKYC¶
| Use Case | How It Works |
|---|---|
| V-KYC speaker verification | Verify the person on video call is who they claim (voice + face) |
| Voice-based authentication | "My voice is my password" — voiceprint matching |
| Voice anti-spoofing | Detect recorded, synthetic, or cloned voice |
| Accessibility | Voice-guided eKYC for visually impaired users |
Voice Anti-Spoofing (ASVspoof)¶
| Attack | Method |
|---|---|
| Replay | Play recorded voice of victim |
| TTS (Text-to-Speech) | Generate victim's voice synthetically |
| Voice cloning | Clone voice using AI (ElevenLabs, VALL-E) |
| Voice conversion | Transform attacker's voice to sound like victim |
Key Takeaways¶
Summary
- Voice biometrics adds a second modality to face-based eKYC
- Voice cloning (AI-generated voice) is a growing threat — parallels deepfake for face
- Most practical in Video KYC where audio is already captured
- Key providers: Nuance (Microsoft), ID R&D, Pindrop