Your high-performance gaming mouse can be used as a microphone to spy on you, and it’s the AI’s fault

If using ChatGPT to do evil already cries out to heaven, what a team of researchers from the University of California has discovered is even more alarming. After some tests, this team has revealed a surprising new vulnerability in a device that millions of people use every day: the computer mouse. His technique, dubbed Mic-E-Mouse, shows that it is possible turn it into a microphone to listen to private conversations, using your own motion sensors and the power of artificial intelligence.

The attack takes advantage of the extreme sensitivity of high-performance mice, especially those designed for gaming. The High sensitivity sensors (20,000 DPI or more) Not only do they detect movement on the mat, but they are also capable of picking up the tiny acoustic vibrations that our voices transmit through the desk surface. These vibrations, although imperceptible to us, are registered by the mouse sensor.

It is important to note that, although these very precise sensors are not so common, Yes, it is easy to find them in high-end mice. In any case, a conventional mouse usually has at most about 3000 DPI.

From vibrations to words: the key role of AI

The attack method does not require a complex virus. It would be enough for the victim’s computer to run software, which could be anything from an open source application to a video game. This software would collect raw mouse sensor data, something many applications do legitimately, and send it to an external server for processing.

This is where artificial intelligence comes into play. The raw data, which is nothing more than noise, is first processed with digital signal filters. They are then fed into a neural network model specifically trained to convert those vibrations into audible speech. The results of this proof of concept are alarming: the system reached a speech recognition accuracy from 42% to 61%.

Although this is academic research for now, Mic-E-Mouse exposes a worrying new avenue for violating acoustic privacy, a type of attack that was unthinkable before the arrival of such sensitive consumer hardware and such advanced AI models. The researchers’ goal is to raise awareness about this vulnerability so that protective measures can be developed in the future.