Search papers, labs, and topics across Lattice.
This paper introduces LEAKFORGE, a novel framework that transforms the challenge of cross-device accelerometer eavesdropping into a physics-guided data-generation problem. By leveraging the constrained variations in audio-to-accelerometer transfer functions, the authors synthesize a diverse set of accelerometer traces from ordinary speech, effectively modeling key physical phenomena. The resulting eavesdropping model, trained on this synthetic data, successfully generalizes to real traces from previously unseen smartphones, demonstrating its practical applicability in the field of acoustic eavesdropping.
LEAKFORGE reveals that eavesdropping on smartphones via accelerometers can be effectively achieved by synthesizing device-specific traces from ordinary speech, challenging assumptions about device variability.
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.
