Вештачка интелигенција

A Revolutionary Approach to Password Detection: Sound-based Typing

Summary

Researchers from Durham University, the University of Surrey, and Royal Holloway, University of London have developed an innovative way of detecting passwords using the sound produced by a keyboard during typing. Their study highlights that the subtle differences in the […]

Novi način otkrivanja lozinki: Pogađanje putem zvuka tastature

Researchers from Durham University, the University of Surrey, and Royal Holloway, University of London have developed an innovative way of detecting passwords using the sound produced by a keyboard during typing. Their study highlights that the subtle differences in the sounds made by each key unintentionally provide a “side channel” of information.

While investigating side channels may seem clumsy and impractical, it serves a useful purpose in enhancing our understanding. This new technique relies on the acoustic emissions of the keyboard, which not only serve as an easily accessible vector for attacks but also mislead victims into underestimating their importance.

For instance, when entering a password, users often conceal their screen, but they do nothing to hide the sound of the keyboard. Researchers even used real-world attack scenarios, such as eavesdropping on laptop keystrokes through a smartphone’s microphone or recording sound during Zoom calls.

This cutting-edge technique utilizes deep learning, an artificial intelligence method that mimics the human learning process, to identify pressed keys on a keyboard. The algorithm was trained on 25 keystrokes for each key of an Apple laptop, using different fingers and pressures. The sound of key presses was meticulously processed, converted into images, and then fed into the deep learning algorithm for classification.

This method has demonstrated exceptionally high accuracy in recognizing keyboard presses, even when recorded via Zoom. It achieved a 95 percent accuracy rate on phone-recorded key presses, which is a significant improvement compared to existing classifiers. Implementing the method on Zoom recordings resulted in an accuracy rate of 93 percent, which is also highly impressive.

This groundbreaking password detection technique utilizes advanced technologies to enhance cybersecurity and raise awareness about potential vulnerabilities. However, the researchers note that further research is necessary to fully understand this field and develop adequate protection methods.

FAQ