Beware! New AI tool can steal passwords by listening to keystrokes

A new AI tool has the capability to identify passwords from sounds of keystrokes captured during Zoom sessions or any other application that has access to your microphone, with an accuracy rate of approximately 90%.

Recent research has revealed that typing your password while on a Zoom call could put you at risk for a cyberattack as hackers can now ‘listen’ to your keystrokes to steal your passwords using an artificial intelligence (AI) tool.

The study revealed that artificial intelligence tools have the ability to identify specific keystrokes by listening to the typing sounds.

The research, conducted by researchers from Durham, Surrey, and Royal Holloway universities, illustrated that microphones can identify distinct typing patterns.

The paper published on August 3, revealed that with the increasing adoption of video conferencing platforms like Zoom and the widespread availability of devices equipped with microphones, the risk of sound-based cyberattacks has increased.

The AI tool studied by the researchers showed the capacity to accurately decipher text, including passwords, from keystroke sounds, achieving a precision level exceeding 90%.

The team of researchers reached this conclusion after undertaking the task of training an AI model.

They achieved this by pressing each of the 36 keys on a MacBook Pro 25 times while recording the accompanying sounds. 

Subsequently, they fed this sound data into the AI model, enabling it to accurately recognise the unique pattern associated with each key.

The researchers’ AI model exhibited a remarkable accuracy rate of 93% in interpreting keystrokes from recorded MacBook typing sounds conducted via the Zoom video conferencing software. 

Additionally, the accuracy rate increased to 95% when the keystrokes were captured using an iPhone 13 mini.