This is a simple version of the Deep Fingerprinting (DF) attack by Sirinam et al.. It is more-or-less easy to use, and limited to the open world setting with distinct labels for monitored websites (and one for all unmonitored websites).
Running df_train.py takes significant time and is ideally used with access to a decent
GPU for deep learning. The result of training is a trained model (for the model itself,
see df_model.py). The model is used to predict (classify) traffic traces using
df_predict.py. The result of prediction is a file with the resulting predicitions.