Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HVAC-EAR: Eavesdropping Human Speech Using HVAC Systems

Tarikul Islam Tamiti, Biraj Joshi, Rida Hasan, Anomadarshi Barua

TL;DR

HVAC-EAR reconstructs intelligible speech from low-resolution, noisy pressure data with two key contributions, surpassing prior work limited to hot word detection and mitigating transient HVAC noise.

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.

HVAC-EAR: Eavesdropping Human Speech Using HVAC Systems

TL;DR

HVAC-EAR reconstructs intelligible speech from low-resolution, noisy pressure data with two key contributions, surpassing prior work limited to hot word detection and mitigating transient HVAC noise.

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.

Paper Structure

This paper contains 18 sections, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Physics, Range and Sampling Frequencies of DPSs
  4. Attack Model
  5. HVAC-EAR Architecture Design
  6. Complex Encoders and Decoders
  7. Complex Skip Block and Complex Conformer
  8. Complex Unified Attention Block (CUAB)
  9. Complex Multi-Resolution STFT Loss
  10. Data Collection and Evaluation
  11. Data Collection and Evaluation
  12. Data Collection from a Real-World Facility
  13. Comprehensive Evaluation Metrics
  14. Comparison with Other Models
  15. Subjective Analysis
  16. ...and 3 more sections

Figures (4)

  • Figure 1: (Left) Internals of a DPS. (Right) An overview of the attack model. DPSs are positioned close to human occupants.
  • Figure 2: (Left) Architecture of HVAC-EAR. (Middle) Details of CUAB. (Right) Real-world data collection and evaluation.
  • Figure 3: (Left) Evaluation using BMS and DPSs. (Right) Reconstructed speech from noisy pressure data of 3.5 dB SNR.
  • Figure 4: (Left) MOS. (Right) Impact of speaker distance.