Trained with keystrokes on a laptop transmitted over a smartphone, a new AI model was able to overhear typing and steal passwords with 95% accuracy.

A team of researchers at Cornell University presented their report “A Practical Deep Learning-Based Acoustic Side Channel Attack on Keyboards” this month, which in addition to a smartphone, trained the AI model over conferencing app Zoom. With the latter, the model was able to replicate passwords with 93% accuracy.

“With recent developments in deep learning, the ubiquity of microphones and the rise in online services via personal devices, acoustic side channel attacks present a greater threat to keyboards than ever,” the researchers explained. “Our results prove the practicality of these side channel attacks via off-the-shelf equipment and algorithms.”