Skip to content
NewA calmer, faster ZeroHour
Researchlow

Pop Goes the Verifier poisons voice liveness detection

Original title: Pop Goes the Verifier: Phoneme-Targeted Data Poisoning of Voice Liveness Detection

arXiv cs.CRby Eshaq Jamdar ()1 min read
Read at arxiv.org
AI summary · x-ai/grok-4.7Summaries by Vani AI

Phoneme-targeted poisoning using FastSpeech 2 degrades ASVSpoof 2019 voice liveness detection accuracy with stealth against waveform inspection.

Pop Goes the Verifier injects localized pop bursts at plosive phonemes using FastSpeech 2 with modified HiFi-GAN vocoder and Gaussian noise. The attack reduces ASVS accuracy while resisting signal-level inspection better than broad SyntheticPop. It achieves moderate poisoning on the ASVSpoof 2019 dataset.

Key facts
  • Injects pop bursts at plosive phoneme locations
  • Uses modified HiFi-GAN vocoder
  • Resists waveform and spectrogram inspection
  • Tested on ASVSpoof 2019
Security assessment · AI, from the articleNo known exploitation

Impact: degrades voice liveness detection accuracy

Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat
Tags
voicepoisoningasvliveness
Products
Automated Speaker Verification Systems
AI models
FastSpeech 2HiFi-GAN

From the original article

1 min read at arxiv.org

Many Automated Speaker Verification Systems (ASVs) commonly employ liveliness detection techniques, such as VoicePop, which identify genuine speech by analyzing low- frequency pop bursts from plosive phonemes using Gammatone Frequency Cepstral Coefficients (GFCC) and an SVM classifier. While effective against traditional replay attacks, such systems are vulnerable to data poisoning techniques.

arXiv cs.CR

By Eshaq Jamdar · arxiv.org

Read the full article at arXiv cs.CR

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.