A study says, AI tools can crack passwords with 93% accuracy by listening to keystrokes

Study warns of cyberattack risk during Zoom calls due to AI's ability to 'listen' to keystrokes, stealing passwords.

AI tool deciphers typing sounds, achieving over 90% accuracy in extracting text, including passwords.

Researchers from Durham, Surrey, and Royal Holloway universities reveal microphones can detect typing patterns, posing a threat in public settings.

AI model trained by recording MacBook Pro keystrokes achieves 93% accuracy on Zoom recordings.

Accuracy improves to 95% when keystrokes are captured using an iPhone 13 mini.

Researchers pressed 36 keys 25 times, feeding sound data into AI model for recognition.

Widespread video conferencing adoption and microphone-equipped devices raise sound-based cyberattack risk.

Paper published on August 3 highlights rising threat as AI gains ability to interpret typing sounds.

Zoom and similar platforms increase vulnerability to sound-based attacks due to microphone accessibility.

Public laptop users particularly susceptible to having keystrokes captured and decrypted by hackers.