High-performance mice can be used as a microphone to spy on users (www.tomshardware.com)

Researchers at UC Irvine published a proof-of-concept called “Mic‑E‑Mouse” showing that high-performance optical mice can be repurposed as covert microphones. Modern gaming mice with very high polling rates (≥8 kHz) and sensitive motion sensors pick up minute acoustic vibrations on the desk surface; by harvesting raw high-rate mouse packets from a compromised host (which can be as simple as benign-looking software or a web app that requests high-frequency mouse data), attackers can reconstruct audio. The team applies digital signal processing (a Wiener filter) to denoise the signal and then a neural model to further enhance speech, achieving automatic speech recognition rates of roughly 42–61% on their tests. This is significant because it turns a ubiquitous, inexpensive peripheral into an unexpected side-channel, enabled by sensor sampling rates rather than DPI (DPI is not the causal factor—polling frequency matters due to Nyquist limits). The attack underscores new privacy risks as AI models improve reconstruction from weak signals. Practical mitigations include restricting high-frequency mouse polling at OS or driver level, limiting web/APP access to raw input streams, firmware/driver integrity checks, and telemetry to detect unusual packet collection. The paper is a warning shot: inexpensive hardware plus ML can create novel acoustic side-channels that security tooling and permission models must address.