Wi-Fi Sensing in the Wild
Wi-Fi networks are everywhere – but beyond connectivity, they can also sense.
In this project, we rethink Wi-Fi sensing by leveraging standard-compliant MU-MIMO beamforming feedback (BFI/BFA) rather than proprietary or firmware-hacked CSI.
Our work spans four major contributions:
- Wi-BFI: The first open-source tool to extract BFAs/BFI in the wild
- Feedback Quantization & Grouping: A benchmark study on how IEEE 802.11 feedback compression affects performance
- BFA-Sense: A learning framework for sensing directly from BFAs, no firmware hacks required
- BeamSense: A robust, cross-domain few-shot sensing system that outperforms CSI-based methods

CSI vs proposed Beamforming Feedback based sensing approach


Left: Wi-BFI can concurrently extract BFAs from devices operating with different standards, channels, and bandwidths. Right: BeamSense, a new approach to Wi-Fi sensing where the standard-compliant BFAs routinely sent in MU-MIMO Wi-Fi networks is used to characterize the propagation environment between the MU-MIMO users and the AP
Unlike CSI-based approaches, which require specialized tools and only capture single-user links,
BFA-based sensing allows us to capture all MU-MIMO channels simultaneously, dramatically improving spatial diversity.
We demonstrated that sensing with BFAs achieves 10% higher accuracy than CSI approaches, and with our cross-domain adaptation (FAMReS), accuracy improves by up to 30% on unseen environments


Left: BeamSense (BFA) vs. SignFi (CSI) and Impact of the spatial diversity of the beamformees on sensing performance Right: Comparative analysis of BeamSense with unseen environments, and subjects
Publications
- Haque, K.F., Meneghello, F. and Restuccia, F., 2023, October. Wi-BFI: Extracting the IEEE 802.11 beamforming feedback information from commercial Wi-Fi devices. In Proceedings of the 17th ACM Workshop on Wireless Network Testbeds, Experimental evaluation & Characterization (pp. 104-111).
- Feedback Quantization & Grouping – IEEE WCL 2024
- BFA-Sense – IEEE PerCom Workshops 2024
-
BeamSense – Computer Networks 2025