Room for Error: Large-Scale Simulation of Over-the-Air Acoustic Attacks

2026-06-26

AI Summary

This work addresses the limited physical-world transferability of existing adversarial attacks on voice-controlled systems, which often neglect critical real-world acoustic factors such as geometric configuration and detectability. To bridge this gap, the authors propose a Dual-Form Signal-to-Noise Ratio (SNR) metric that decouples attack stealth from effectiveness and develop a high-throughput acoustic simulation platform validated with real-environment testing. Over 8 million adversarial evaluations were conducted on mainstream speech recognition models, including Whisper and wav2vec. Experimental results demonstrate that incorporating realistic acoustic conditions can increase relative word error rates by up to 94.5%, underscoring the profound impact of environmental factors on system security. This study advances acoustic adversarial research toward greater reproducibility and empirical verifiability.

Abstract

While voice control is rapidly becoming a ubiquitous vector of human-AI communication, the risks facing these systems remain poorly understood. This is, in part, a product of the difficulties in scaling strictly digital adversarial workflows to the physical world. These scale barriers have led the community to abstract away key acoustic factors relating to detectability and the influence of geometry on acoustics. These methodological and metrological shortcomings undermine our understanding of risk. We illuminate these issues through real-world testing, conceptual discussions, and a novel, high-throughput reality simulation framework. By testing over 8 million adversarial evaluations, we demonstrate that acoustic awareness yields relative Word Error Rate increases of up to 94.5\% under Whisper and wav2vec. We employ this framework to explore a formalize and operationalize a Dual-Form Signal to Noise Ratio to decouple source stealth from victim attack efficacy, resolving a crucial limitation in current works. This lays the groundwork for repeatable, verifiable research that embraces, rather than abstracts, the acoustic environment.

Problem

• adversarial attacks
• acoustic environment
• voice control systems
• Word Error Rate
• signal-to-noise ratio

Innovation

• acoustic adversarial attacks
• high-throughput simulation
• Dual-Form SNR
• Word Error Rate
• physical-world AI security