AI logs keystrokes by sound alone, putting noisy mechanical keyboard users at greatest risk

Even the Silent Reds are decipherable through artificial intelligence, according to a new paper by a team of researchers from British universities. Their paper on acoustic side channel attack, released last week, says AI can identify keystrokes with 95% accuracy through sound alone.

In the study, experimenters correctly identified keystrokes on a MacBook Pro, overheard through a nearby phone, 95% of the time.

Advertisers from Lebal Drocer, Inc. have already begun using the new technology to learn more about their customers through keystrokes than they ever learned overhearing conversations through the microphone about toilet paper.

Chief researcher at the Lebal Drocer Institute of Consumer Studies, Albert H. Troudemaeier, said he was able to get his colleagues’ passwords during a Zoom meeting.

“No matter the context, if there’s a keyboard singing, this software knows the tune,” Troudemaeier said. “With recent developments in microphone technology, as well as deep learning models, the rate at which we can determine what our customers want, need — what they fear — has expanded by analyzing the very content of their keystrokes, enabling us to serve them better than we ever could before. It’s very powerful, and uses existing hardware access everyone has already agreed to it in the terms of service.”

Laptops are ideal vectors for analysis because of their portability. People take their laptops to work in public spaces like libraries, whorehouses, and university lecture halls, where the sound of typing is recorded, unnoticed, by every other laptop in the room.

“You can hide your screen,” Troudemaeier said, “but you can’t hide that unmistakable sound. We will find you.”