How can a neural network steal your password using a microphone?

A group of researchers from leading British universities has identified a new, potentially dangerous method of stealing passwords through the analysis of the sound of keystrokes on the keyboard. The news once again highlights the importance of data security in the digital age, according to Bleeping Computer.

The researchers have developed a neural network model that is able to recognize the sounds of pressing each key on the keyboard with an accuracy of 95%. To do this, they recorded the sounds of the keys using a microphone. This method was able to successfully recognize the unique sounds made by different keys on an Apple MacBook Pro laptop.

However, when using this method in the conditions of the video conferencing service Zoom, the accuracy of the model decreased slightly, up to 93%. However, this still means that hackers can pick up almost any password the user enters.

The peculiarity of this attack is that it does not require complex conditions or special equipment. Modern devices ranging from laptops to smartphones already have high quality microphones, making this method more accessible to attackers.

Experts advise using additional authentication methods, such as fingerprint scanning or other biometric mechanisms, to provide an additional layer of protection against such “sonic” attacks on passwords.