Periscope: A Keystroke Inference Attack Using Human Coupled Electromagnetic Emanations
View/ Open
Date
2021-11-19Author
Wenqiang, Jin
Srinivasan, Murali
Zhu, Huadi
Li, Ming
Metadata
Show full item recordAbstract
This study presents Periscope, a novel side-channel attack that
exploits human-coupled electromagnetic (EM) emanations from
touchscreens to infer sensitive inputs on a mobile device. Periscope
is motivated by the observation that finger movement over the
touchscreen leads to time-varying coupling between these two.
Consequently, it impacts the screen’s EM emanations that can be
picked up by a remote sensory device. We intend to map between
EM measurements and finger movements to recover the inputs.
As the significant technical contribution of this work, we build an
analytic model that outputs finger movement trajectories based on
given EM readings. Our approach does not need a large amount of
labeled dataset for offline model training, but instead a couple of
samples to parameterize the user-specific analytic model. We implement Periscope with simple electronic components and conduct a
suite of experiments to validate this attack’s impact. Experimental
results show that Periscope achieves a recovery rate over 6-digit
PINs of 56.2% from a distance of 90 cm. Periscope is robust against
environment dynamics and can well adapt to different device models and setting contexts.