Recently, researchers at the University of California, Irvine, unveiled their latest findings. By harnessing the subtle vibrations captured by high-performance mouse sensors and leveraging AI tools to transform these vibrations into sound data, the mouse can effectively function as a listening device akin to a microphone. Dubbed "Mic-E-Mouse," this project highlights that mouse sensors boasting a sensitivity of 20,000 DPI or above are susceptible to such eavesdropping techniques. With the widespread adoption of PAW3395/3399 sensors, the number of users potentially vulnerable to these attacks has surged.

However, it's important to note that the prerequisite for such an attack is that the user connects their mouse to a compromised computer. Infected open-source software or even normal software, if compromised, could serve as potential attack vectors. Current tests indicate that the Mic-E-Mouse achieves a speech recognition accuracy rate ranging from 42% to 61%. As AI technology continues to evolve and mature, it is anticipated that this accuracy rate will see further enhancements.