PhaseShield: Adversarial Perturbation of WiFi Channel State Information for Privacy-Preserving Passive Sensing Mitigation

WiFi Channel State Information (CSI) sensing enables passive, contactless reconstruction of human presence, skeletal pose, and vital signs from ambient 802. 11 OFDM signals using commodity hardware costing under 50, with no authentication required and no modification to the target network. Despite the rapid maturation of CSI-based sensing -achieving throughwall pose estimation and sub-BPM vital sign accuracy in recent work -the defense literature remains almost entirely absent. This paper presents PhaseShield, the first systematic framework for disrupting WiFi CSI-based passive sensing through adversarial perturbation at the physical signal layer. We establish a formal signal model for the human-induced multipath channel and decompose the sensing pipeline into three exploitable properties: subcarrier phase coherence, temporal frame stability, and Doppler-shift isolation. We derive four perturbation algorithms -PhaseShield-FGSM, -PGD, -Random, and a subcarrier-selective variant -each targeting one or more of these properties while remaining within the OFDM equalization correction range of legitimate receivers, preserving network function. A subcarrier sensitivity analysis method identifies the privacy-critical subcarrier set Ψ: a small minority (10-15%) of subcarriers that contribute disproportionately to sensing confidence. Simulation results using the opensource RuView production pipeline predict detection confidence collapse from 0. 72-0. 89 (operational) to below 0. 15 (below threshold) at ε ≤ 0. 3 rad. We present three patentable claims, an IEEE 802. 11 amendment proposal providing router-level CSI privacy, and a comprehensive analysis of adaptive adversaries. We further present a hardware prototype design based on the ESP32-S3 microcontroller, detailing the system architecture, power budget, and three concrete deployment scenarios.