“Mic-E-Mouse” Attack Lets Hackers Steal Sensitive Data via Mouse Sensors

A new cybersecurity vulnerability called the “Mic-E-Mouse” attack turns high-performance optical sensors in consumer mice into covert eavesdropping devices, secretly capturing confidential conversations through acoustic vibrations on work surfaces. According to researchers, the attack uses signal processing and machine learning to reconstruct intelligible speech from the captured signals, despite initial poor quality and sampling challenges.

In tests with consumer-grade sensors, the method achieved a SI-SNR increase of +19dB, an 80% speaker recognition accuracy in automated tests, and a human study word error rate of 16.79%. The attack can capture human speech frequencies between 200Hz and 2000Hz, and the researchers note that high-performance mice are increasingly accessible, with open-source devices available for under £50, broadening the potential impact across personal, corporate, and government environments.

The threat model focuses on open-source applications where high-frequency mouse data collection seems legitimate, and delivery vectors include creative software and video games that repurpose networking code to discreetly exfiltrate data. The discovery highlights a previously unknown vector that could turn everyday peripherals into invisible surveillance devices. Oct 8 2025.