Wireless localization systems rely on channel state information (CSI) to determine position, but physical-layer spoofing attacks can tamper with that data. Such attacks pose a serious threat, as manipulated CSI can cause a system to trust a false location.
The paper introduces an adaptive Gaussian mixture model for anomaly detection at the physical layer. The model is designed to identify deviations in CSI that indicate spoofing, with an adaptive component intended to keep up with changing channel conditions.
The authors note that detection is especially challenging when the attacker and the legitimate source are hard to distinguish. The abstract highlights this difficulty, though full results are not included in the excerpt.