"New 'Deep Learning Attack' Deciphers Laptop Keystrokes with 95% Accuracy"

A group of researchers has developed a "deep learning-based acoustic side-channel attack" that is 95 percent accurate in classifying laptop keystrokes recorded by a nearby phone. According to the researchers Joshua Harrison, Ehsan Toreini, and Maryam Mehrnezhad, when trained on keystrokes recorded with the video conferencing software Zoom, an accuracy of 93 percent was reached, a new record for the medium. Side-channel attacks are a class of security exploits aimed at gaining information from a system by monitoring and measuring its physical effects while processing sensitive data. Typical observable effects include runtime behavior, power consumption, electromagnetic radiation, acoustics, and cache accesses. To execute the attack, the researchers first conducted experiments with 36 of the Apple MacBook Pro's keys (0-9, a-z), pressing each key 25 times in a row, varying in pressure and finger. The next step involved isolating the individual keystrokes and converting them into a mel-spectrogram, on which a deep learning model called CoAtNet was run to classify the keystroke images. This article continues to discuss the deep learning-based acoustic side-channel attack. 

THN reports "New 'Deep Learning Attack' Deciphers Laptop Keystrokes with 95% Accuracy"