Abstract: In recent years, concerns are increasingly being expressed about the threats of side-channel attacks that exploit acoustic emanations from electronic as well as mechanical devices of daily use. With the increased level of sophistication in both hardware and applications that run on mobile phones, the number of possible ways in which their vulnerabilities can be exploited is also on the rise. In this article, we demonstrate a novel attack which uses the sound emanating from a tap made on the touchscreen of a smartphone to decipher the text being typed. The audio signal captured by the pair of microphones typically embedded in a smartphone is first processed to determine a candidate set of keys. Filters are employed to make this step robust against ambient noise. Natural language processing techniques are then used to estimate the most probable words and sentences that can be constructed from a sequence of taps. It is shown that using even off-the-shelf tools, the typed text including passwords can be guessed with reasonably high accuracy. Besides raising awareness about this potential side-channel attack, we identify the causes that allow it to succeed and suggest countermeasures.
