How to Get Away with MoRTr: MIMO Beam Altering for Radio Window Privacy

We consider the radio window attack, a privacy threat in which an attacker monitors the wireless link over a period of time, recording the channel state information (CSI) across multiple packets and uses a model to detect, estimate, or classify human movements. To prevent such privacy attacks on a wireless channel, we propose modifying radio training (MoRTr), a novel system for Wi-Fi MIMO-OFDM devices that alters transmitted symbols over time, space and frequency via a pseudo-random process that mimics the changes due to human activity, particularly the training symbols that are used to measure the wireless channel by the receiver. We perform extensive experiments to demonstrate that an attacker is thwarted by the approach. At the same time, we demonstrate that any receiver is able to use its measured CSI to demodulate the data without any significant degradation in performance, despite the fact that the receiver is not measuring the true CSI.