"Silent Data Theft: Acoustic Attacks Capture Keystrokes with 95% Accuracy"

Researchers have developed a deep learning model called CoAtNet that can steal data from keyboard keystrokes recorded using a microphone with an accuracy of 95%. Even when using Zoom for training, the accuracy dropped to 93%. Acoustic attacks have become simpler due to the abundance of microphone-bearing devices, posing a significant threat to data security. Possible mitigations include altering typing styles, using randomized passwords, employing biometric authentication, and utilizing password managers.