With an accuracy rate of around 90%, a new AI tool may identify passwords from keystroke noises gathered during Zoom sessions or any other programme that can access your microphone.
According to new research, inputting your password while on a Zoom conversation puts you in danger of a cyberattack since hackers can now use an artificial intelligence (AI) application to ‘listen’ to your keyboards and collect your passwords.
According to the findings, artificial intelligence programmes can recognise certain keystrokes by listening to typing sounds.
The study by Durham, Surrey, and Royal Holloway universities demonstrated that microphones can detect various typing patterns.
According to the rep
The researchers’ AI model was 93% accurate in recognising keystrokes from recorded MacBook typing noises conducted over Zoom video conferencing software.
Furthermore, when the keystrokes were collected using an iPhone 13 mini, the accuracy rate climbed to 95%.
ort, which was released on August 3, the risk of sound-based cyberattacks has increased due to the increasing popularity of video conferencing services like Zoom and the ubiquitous availability of devices equipped with microphones.
The researchers’ AI programme demonstrated the ability to reliably decipher the text, including passwords, from keystroke sounds, with a precision level above 90%.
The researchers arrived at this result after completing the task of training an AI model.
They accomplished this by tapping each MacBook Pro’s 36 keys 25 times while recording the associated sounds.
They then fed this sound data into the AI model, allowing it to distinguish the distinct pattern associated with each key.