Recovering keystrokes from audio recordings of keyboard emanations using time difference of arrival data and an English language model

Sam Thornton


Supervised by Kirill Sidorov; Moderated by Alun D Preece

Audio emanations from keyboards have been found to contain information relating to the typed content. Given

several audio recordings of typing, is it possible to recover the text?

Using the difference between wavefronts arriving at microphones to build a geometric model, and a simple model of the English language, how much text can be recovered? This method is based on prior research by Asonov and Agrawal, Zhu et al., and Zhuang et al..

Initial Plan (06/02/2023) [Zip Archive]

Final Report (19/05/2023) [Zip Archive]

Publication Form