Original title: Cross-Domain Inference for Human Localization: Applying Wi-Fi RSSI Data to CSI-Trained Models
Ariel Duschanek-Myers, Thomas Welsh, Helmut Neukirchen · September 2026
Wi-Fi signals can secretly reveal where you are, threatening your privacy. Many tracking systems use detailed signal details called Channel State Information, or CSI. However, normal smart devices cannot easily record CSI because it demands special software and deep system permissions. To see if simpler gadgets pose the same threat, researchers looked at the Received Signal Strength Indicator, or RSSI. RSSI measures basic signal power in decibel-milliwatts, and almost any device can read it without special access. Instead of spending time training a fresh computer model from scratch, the team tested cross-domain inference. They fed simple RSSI readings into an existing system originally built for complex CSI data. By checking the model against video of a person moving in a room, they found it predicted locations with roughly 80% confidence during movement. This reveals that common, everyday smart gadgets can track people and invade privacy.
Imagine trying to guess what your neighbor is cooking from the smell drifting out the window. You have only ever cooked from detailed recipes, so you know exactly what each ingredient looks like, but now you have to work from a single vague scent that is far less precise. In this paper, the recipe is the detailed Wi-Fi data the system was trained on, and the scent is the simple signal strength it is given instead—so the system, trained on the recipe, is asked to make do with the scent, and it turns out it can, roughly.
Most people carry devices that constantly scan Wi-Fi. Until now, accurate location tracking seemed to require special hardware or deep system access. This work shows that ordinary signal strength readings—available to almost any smart device—can locate a moving person with high confidence, meaning widespread privacy invasion becomes much easier than assumed.
The abstract doesn't say whether the approach works for multiple people, stationary people, or different room layouts. It also doesn't specify the false-alarm rate or how the 80% confidence was measured. The evaluation appears to be limited to a single room with one moving person, so generalization to real-world Wi-Fi environments remains untested.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Cross-Domain Inference for Human Localization: Applying Wi-Fi RSSI Data to CSI-Trained Models — Ariel Duschanek-Myers, Thomas Welsh, Helmut Neukirchen
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.