A report reveals that researchers designed an AI machine to listen to a MacBook Pro key password entered over a cell phone and Zoom conversation.
The AI system immediately recognized each key sound. After that, the algorithm could detect each key sound with a precision of 95% over the cell phone and 93% using Zoom.
The classifier had the best accuracy without a language model, 95%, while trained on keystrokes from a nearby phone.
During a Zoom chat, the hack-friendly artificial intelligence tool was accurate while “listening” to the typing over the microphone of the laptop.
According to the researchers, it was able to duplicate the keystrokes with an accuracy of 93%.
The researchers issued a warning that a lot of users are unaware of the potential that malicious actors may observe their typing in order to breach accounts. They referred to this form of hack as an acoustic side-channel attack.
For instance, while a person is inputting a password, they will often cover the screen but do very little to obscure the sound of their keyboard.
The accuracy of the measurement was determined by the researchers by pressing each of the laptop’s 36 keys a total of 25 times, with varied degrees of pressure and different fingers.
The software was able to “listen” to different aspects of every key press, including the sound wavelengths, and detect those aspects. A distance of 17 millimeters was maintained between the keyboard and the smartphone, an iPhone 13 mini.
Researchers Ehsan Toreini from the University of Surrey, Joshua Harrison from Durham University, and Maryam Mehrnezhad from the Royal Holloway University of London were the ones who carried out the study.
A report shows another potential threat posed by the rapidly developing technology is the idea that AI technologies may assist cyber criminals.
Without the appropriate safeguards in place, artificial intelligence (AI) might represent a substantial threat to mankind, according to several prominent industry professionals, including OpenAI founder Sam Altman and billionaire Elon Musk.