This thesis investigate if applications can be identified from publicly avialable meta data, from a Wi-Fi network.

The classification was done through clustering by a Gaussian mixture model -- on data with restricted labelling.
The data was captured at Aalborg University by passively monitoring Wi-Fi traffic.
By utilising three different preprocessing methods -- Simple standardisation, transformation through Factor analysis of mixed data and a trained Gated recurrent unit - autoencoder, encoding -- features of the data is transformed for the clustering.

Even with the used preprocessing methods, the clustering were found no better than random guessing for identifying applications for the captured data labelling combination.