Using Unrestricted Mobile Sensors to Infer Tapped and Traced User Inputs

摘要

As of January 2014, 58 percent of Americans over the age of 18 own a smart phone. Of these smart phones, Android devices provide some security by requiring that third party application developers declare to users which components and features their applications will access. However, the real time environmental sensors on devices that are supported by the Android API are exempt from this requirement. We evaluate the possibility of exploiting the freedom to discretely use these sensors and expand on previous work by developing an application that can use the gyroscope and accelerometer to interpret what the user has written, even if trace input is used. Trace input is a feature available on Samsung's default keyboard as well as in many popular third-party keyboard applications. The inclusion of trace input in a key logger application increases the amount of personal information that can be captured since users may choose to use the time-saving trace-based input as opposed to the traditional tap-based input. In this work, we demonstrate that it is indeed possible to recover both tap and trace inputted text using only motion sensor data.