HVAC-EAR: Eavesdropping Human Speech Using HVAC Systems

2025-10-01

AI Summary

This work identifies a novel voice eavesdropping threat posed by pressure sensors in HVAC systems due to their unintended sensitivity to acoustic pressure. To reconstruct intelligible speech from low-sampling-rate (as low as 0.5 kHz), high-noise, and low-resolution pressure signals, we propose a complex-valued Conformer-based speech reconstruction framework. It introduces a complex unified attention block to model phoneme-level temporal dependencies and jointly recovers spectral magnitude and phase—effectively suppressing HVAC-induced transient noise—enabling end-to-end complex-domain spectral reconstruction. Evaluated on real-world pressure measurements collected from operational HVAC environments, our method achieves the first demonstration of highly intelligible speech reconstruction at 0.5 kHz sampling (reducing ASR word error rate by 37%). This result empirically exposes a previously overlooked covert voice leakage vulnerability in existing building infrastructure and provides both critical evidence and a practical technical countermeasure for physical-layer privacy security.

Technology Category

Natural Language Processing: SpeechMachine Learning: PrivacyData Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

Security and Privacy: Large-scale security measurementsSystems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environmentsUser Modeling, Personalization and Recommendation: User privacy protection in personalized systems

Abstract

Pressure sensors are widely integrated into modern Heating, Ventilation and Air Conditioning (HVAC) systems. As they are sensitive to acoustic pressure, they can be a source of eavesdropping. This paper introduces HVAC-EAR, which reconstructs intelligible speech from low-resolution, noisy pressure data with two key contributions: (i) We achieve intelligible reconstruction from as low as 0.5 kHz sampling rate, surpassing prior work limited to hot word detection, by employing a complex-valued conformer with a Complex Unified Attention Block to capture phoneme dependencies; (ii) HVAC-EAR mitigates transient HVAC noise by reconstructing both magnitude and phase of missing frequencies. For the first time, evaluations on real-world HVAC deployments show significant intelligibility, raising novel privacy concerns.

Problem

• Reconstructs intelligible speech from HVAC pressure sensor data
• Achieves speech reconstruction from extremely low sampling rates
• Mitigates transient HVAC noise by reconstructing frequency components

Innovation

• Uses complex-valued conformer for speech reconstruction
• Employs unified attention to capture phoneme dependencies
• Reconstructs both magnitude and phase of frequencies