Reproduce “Your Microphone Array Retains Your Identity: A Robust Voice Liveness Detection System for Smart Speakers” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yan Meng, Jiachun Li, Matthew Pillari, Arjun Deopujari, Liam Brennan, Hafsah Shamsie, Haojin Zhu, Yuan Tian

Researchers from Shanghai Jiao Tong University and the University of Virginia developed ARRAYID, a voice liveness detection system for smart speakers that exploits microphone arrays to create a robust "array fingerprint." This novel approach achieved 99.84% accuracy with a 0.17% Equal Error Rate (EER) on a new multi-channel dataset, significantly outperforming previous passive methods and demonstrating strong resilience to environmental variations and advanced spoofing attacks.

Key insights

  • The standard deviation of sound propagation across a microphone array (the array fingerprint) exhibits stability and distinct characteristics between live and spoofed voices, irrespective of the audio source's position or environmental changes, particularly with four or more microphones.
  • Exploiting the multi-channel nature of microphone arrays allows for effective cancellation of environmental noise and distortions, providing a more reliable liveness feature than single-channel approaches.
  • A combination of the novel array-derived features with traditional audio features (like LPCCs) enables a comprehensive and robust defense against various voice spoofing methods, including sophisticated adversarial attacks.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.