Virginia Tech® home

Scheduled Spatial Sensing against Adversarial WiFi Sensing

Paper Details


WiFi sensing uses changes in Channel State Information (CSI), or how a signal travels from a transmitter to a receiver as they reflect from objects in the environment. Through the utilization of machine learning classification models developed based on this CSI data, WiFi sensing can predict physical actions (for example, when people walk or run) in the environment. 

This technology provides low-cost motion detection and activity recognition in smart homes as it can leverage already available WiFi infrastructure.

These systems have recently been commercialized, and their usage will grow. However, the security aspect of such systems has not been widely considered. WiFi signals can be sniffed by adversaries, and malicious actors can use the same sensing method to learn private information about residents.

Researchers proposed a solution through multiple transmitter antennas linked to a single source, which communicates with a receiver based on a schedule. 

This simple, low-cost system allows legitimate receivers that have the transmitter antenna schedule to recognize the activities more accurately. Eavesdroppers or adversaries, who are unaware of the transmission schedule, fail to perform proper sensing.