KeyAtNet: Keystroke Signal Real-time Eavesdropping Based on Fourier Neural Operator

In this study, we investigate the subtle variations in pitch observed when different keys on a keyboard are pressed. Through the application of Fourier Transform, we analyze the original signals and identify distinct differences in the frequency domain corresponding to various keystrokes. Drawing inspiration from Fourier Neural Operator (FNO), we introduce the Key Attack Network (KeyAtNet) tailored for real-time eavesdropping of keyboard keystroke signals. By leveraging raw audio signals as input to the KeyAtNet model, we enhance the efficiency of real-time inference and ensure the accuracy of eavesdropping.