CSI:DeSpy

Abstract

Recently, the spy cameras spotted in private rental places have raised immense privacy concerns. The existing solutions for detecting them require additional support from synchronous external sensing or stimulus hardware such as on/off LED circuits, which require extra obligations from the user. For example, a user needs to carry a smartphone and laboriously perform preset motions (e.g., jumping, waving, and preplanned walking pattern) for synchronous sensing of acceleration signals. These requirements cause considerable discomfort to the user and limit the practicability of prevalent solutions. To cope with this, we propose CSI:DeSpy, an efficient and painless method by leveraging video bitrate fluctuations of the WiFi camera and the passively obtained Channel States Information (CSI) from user motion. CSI:DeSpy includes a self-adaptive feature that makes it robust to detect motion efficiently in multipath-rich environments. We implemented CSI:DeSpy on the Android platform and assessed its performance in diverse real-life scenarios, namely; (1) its reliability with the intensities of physical activities in diverse multipath-rich environments, (2) its practicability with activities of daily living, (3) its unobtrusiveness with passive sensing, and (4) its robustness to different network loads. CSI:DeSpy attained average detection rates of 96.6%, 96.2%, 98.5%, and 93.6% respectively.