AFTER : A Fine-grained Traffic prEdiction fRamework

run_multiple_experiments_1step.py, the main script to perform an experimental campaing

Input Parameters

--confid (-c) : configuration file to run an experiment (it will be edited automatically by the framework)

--proto (-c) : used to train a model at PROTO level on the specified transport-level protocol (choices={TCP, UDP})

--finter (-f) : enables filtering based on app (a), activity (t) and on both (m) (choiches={a, t, m})

--output (-o) : used to set the folder path to store the final results

--finter (-f) : used to filter traffic based on app, activity, and both of them. If a is passed, it will train a model on the specified APP traffic only (APP-level model). If t is passed, it will train a model on the traffic of the specified ACTIVITY only. If m is passed, it will train a model only on traffic of the specified APP-ACTIVITY combination. This parameter could be used in combination with the app parameter (the next one)

--app (-a) : used to specify the label that should be used to filter traffic, according to the filtering-level. It has to be used only when finter is used

--window (-w) : used to set the window (memory) size (i.e., the number of packets used by the model to provide the predictions)

--gpu-id (-g) : used to pass the ID of the GPU which must be used to run the experiments

--sensitivity (-s) : enables the sensitivity analysis on the window size

--validation (-v) : allows the use of a validation-set to perform training (i.e., early-stopping based on validation-loss, otherwise training-loss is used)

--model (-m) : used to specify the Deep Learning architecture. choice={CNN, LSTM, GRU, SERIES_NET}

--sampling (-S) : enables the subsempling of the entire dataset (if used it uses the 10% of the dataset)

Demo to launch an experiment

python3 run_multiple_experiments_1step.py -c ./config_base.ini -e ./Experiments/ -o ./Experiments/FinalResults/ -v -m CNN -w 10

INFO

Tested on python3.8

The list of dependencies is provided in the requirements.txt file inside the root folder

Acknowledgements

This work is supported by the “ADDITIONAL” Project funded by Vietsch Foundation