Location- verification and network planning via machine learning approaches

We consider the in-region location verification problem of deciding if a message coming from a user equipment over a wireless network has been originated from a specific physical area (e.g., a safe room) or not. The detection process exploits the features of the channel over which the transmission occurs with respect to a set of network acces points. A machine learning approach is used, based on a neural network. The neural network is trained with channel features (in particular, noisy attenuation factors) collected by the acces point for various positions of the user equipment both inside and outside the specific area. By seeing the in-region location verification problem as an hypothesis testing problem, we address the optimal positioning of the acess point for minimizing either the area under the curve of the receiver operatinc characteristic or the cross entropy between the neural network output and its correct label. We propose a two-stage particle swarm optimization algorithm having as target first the minimization of the cross entropy and the receiver operating characteristic area under the curve in the two stages. Through simulations we show that for long training and neural network with enough neurons the proposed solution achieves the performance of the Neyman-Pearson lemma.