A high-performance mouse turns into a bug, and AI can sense desktop vibrations to identify conversations

The University of California, Irvine, recently published the "Mic-E-Mouse" study, which proves that high-performance mouse sensors can be turned into listening devices through AI technology. Researchers Isaac Tunney, John Bass and Alexis Lussier Desbiens discovered that a sensor mouse with a 20,000DPI or above can detect weak acoustic vibrations on the desktop. After signal processing and machine learning analysis, it can even "hear" nearby conversations.

The high-performance mouse sensor can poll up to 8,000 times per second, far exceeding the frequency of ordinary mice. When sound waves propagate through the desktop, the sensor detects transverse vibrations of the desktop. These subtle signals, which are ignored by normal use, contain the frequency range needed to reconstruct the sound. The team first used Wiener Filter to filter background noise, and then used a neural network model to analyze sound waves. The current speech recognition accuracy is 42% to 61%. As AI improves, the accuracy can be improved.

Affordable mouse poses security concern

The team pointed out that high-performance sensors such as PixArt PAW3395 and PAW3399 have lowered the price of high-end mice to US$100~300, greatly increasing the number of potentially vulnerable users. The test mouse is a high DPI sensor model such as the HyperX Haste 2 S.

An attacker can collect mouse data using normal-looking open source software, such as creative applications or video games that require high-frequency mouse data support. Common library programs such as Qt, GTK or SDL can also be used to collect mouse signals with normal user rights without the need to obtain sudo administrator rights. Raw data can be exported from the target computer and post-processed by other devices. This monitor will operate at typical office or home discussion levels of 60 to 80 decibels.

This is reminiscent of a classic case from the Cold War era. The Soviet State Security Council (KGB) implanted a passive microphone into a replica of the national emblem given to the U.S. Embassy in Moscow. It was activated only when it received a high-frequency radio signal. Therefore, it was hidden for nearly seven years before it was discovered. The concept of "Mic-E-Mouse" is similar. It also uses a seemingly harmless device to monitor, but after the rise of AI, the completeness of data interpretation is even more complete than before.

Motion sensors in high-performance mice can be used as a microphone to spy on users, thanks to AI — Mic-E-Mouse technique harnesses mouse sensors, converts acoustic vibrations into speech