Shocking! AI Attains 95% Accuracy In Snatching Passwords Via Keystroke Listening, Study Shows

A recent study has unveiled a concerning revelation: malicious actors could exploit artificial intelligence (AI) tools to effectively pilfer user passwords by surreptitiously capturing the sounds of keystrokes, as indicated by a new investigation.

As outlined in the findings disclosed within the research from US-based Cornell University , the AI software, once activated on a nearby smartphone, astonishingly achieved an impressive 95 percent accuracy in accurately replicating the entered passwords.

Notably, a consortium of computer scientists from the UK undertook the task of training an AI model to discern the auditory cues of keystrokes on the 2021 iteration of a MacBook Pro—referred to as a "popular off-the-shelf laptop", as detailed by the New York Post .

Even during a virtual Zoom video conference, the AI tool, which holds potential for exploitation by hackers, showcased remarkable precision in capturing keystrokes emanating from the laptop's microphone. The researchers documented an impressive 93 percent accuracy rate in reproducing these captured keystrokes, setting a notable record for this mode of operation.

Furthermore, the research team issued a cautionary note, highlighting that a significant number of users remain unaware of the lurking danger wherein malicious entities can clandestinely monitor their typing activities to compromise accounts. This particular form of cyber assault is dubbed an "acoustic side-channel attack".

"The ubiquity of keyboard acoustic emanations makes them not only a readily available attack vector but also prompts victims to underestimate (and therefore not try to hide) their output," underscored the study.

"For example, when typing a password, people will regularly hide their screen but will do little to obfuscate their keyboard's sound," it added.

To meticulously evaluate accuracy, the researchers conducted an exercise wherein they pressed each of the laptop's 36 keys 25 times, introducing variations in finger placement and pressure with each press. This approach enabled the AI programme to "listen" for distinctive attributes of each keystroke, such as unique sound wavelengths.

The smartphone employed in this experiment—an iPhone 13 mini—was positioned at a distance of 17 centimetres from the keyboard, thus showcasing the potential for unauthorised interception even from a moderate separation.