Working on an algorithm for recovering the text you type just by analyzing the keyboard sound captured through the computer's mic (i.e. acoustic eavesdropping). The hard part is doing it without having train data for the keyboard.