Let It Beam: Enabling Selective and Secure CSI-based Sensing via Beamforming

Integrated sensing and communication (ISAC) strategies are key components of next-generation wireless networks, including Wi-Fi systems, enabling new services and supporting network management operations. The main idea behind this integration is that the channel state information (CSI), which is continuously estimated for communication purposes, can be leveraged for environmental sensing, especially with the aid of artificial intelligence (AI) algorithms. However, this capability comes with privacy concerns, as passive eavesdroppers, even when using commercial off-the-shelf (COTS) devices, can estimate the CSI and potentially infer sensitive information. In this paper, we propose a new technique to mitigate this risk, leveraging the potentialities of multiple-input multiple-output (MIMO) systems. We design an obfuscation and de-obfuscation system that conceals the real CSI from eavesdroppers, while enabling trusted devices to reverse the distortion and perform sensing normally. We implemented a prototype of our system through software-defined radios (SDRs) and evaluated the effectiveness of our proposed approach considering a device-free localization task. The results show that the obfuscation makes it unfeasible to perform sensing at unauthorized devices (the accuracy drops to about 20%) while the de-obfuscation at legitimate devices allows reaching almost 100% in sensing accuracy. To support reproducibility, we make the dataset and code publicly available.