From Profiling to Parameterization: Physics-Guided Acoustic Eavesdropping via Smartphone Accelerometers

2026-07-28

AI Summary

This work addresses the cross-device acoustic information leakage risk inherent in smartphone accelerometers, a vulnerability exacerbated by the limited generalizability of existing approaches that rely on device-specific modeling. To overcome this limitation, the authors propose LEAKFORGE, a novel framework that, for the first time, parameterizes the physical audio-to-accelerometer transfer mechanism. LEAKFORGE explicitly models key physical effects—including electromechanical coupling, structural resonance, filtering, and aliasing—to construct a family of generalizable transfer functions. Leveraging this parametric model, the framework synthesizes diverse cross-device accelerometer signals without requiring real-device data collection. Consequently, it enables zero-shot transfer to unseen devices and successfully reconstructs high-fidelity speech, thereby demonstrating both the feasibility of cross-device eavesdropping and the strong generalization capability of the proposed approach.

Abstract

We present LEAKFORGE, a device-agnostic framework that converts cross-device accelerometer eavesdropping into a physics-guided data-generation problem. Crucially, device-specific leakage is not arbitrary; its dominant variation lies within a constrained family of audio-to-accelerometer transfer functions. LEAKFORGE samples this family to synthesize large-scale, device-diverse accelerometer traces from ordinary speech, explicitly modeling electromechanical transfer, structural resonances, filtering, and aliasing. An eavesdropping model trained entirely in this synthetic domain can then be applied directly to traces from previously unseen smartphones.

Problem

• acoustic eavesdropping
• smartphone accelerometers
• cross-device
• transfer functions
• device-agnostic

Innovation

• physics-guided
• accelerometer eavesdropping
• transfer function
• synthetic data generation
• device-agnostic