Use the following instructions to obtain the training set to evaluate classification learning algorithms in Weka:

1. Provide the path to the dataMatrix.mat variable file exported by CreateHist() in the Classification.m MATLAB script to create the training set (features + labels).

2. Run Classification.m to create and export the Training set instances in a CSV file. Then run:

• For device classification Python script Dataset_Dev_Classification.py to form the final Training set for Weka.
• For device type classification Python script Dataset_DevType_Classification.py to form the final Training set for Weka.

Examples

1. Example for creating dataset for device classification as in the paper.

• Export the dataMatrix for pre-processed (cleaned) TCP traces:

CreateHist('~/isolatedTestbedData/', 'TCP', 0, 500, 0, 0.1), or use the one that was exported in KD step 3.

• Create the training instances for the training set: run Classification.m
• Form the final training set for Weka: run Dataset_Dev_Classification.py
• Feed the obtained training set into Weka, select the learning algorithm to be trained and cross-validation to evaluate its performance.

2. Example for creating dataset for device type classification as in the paper.

• Export the dataMatrix for pre-processed (cleaned) TCP traces:

CreateHist('~/isolatedTestbedData/', 'TCP', 0, 500, 1, 0.1)

• Create the training instances for the training set: run Classification.m
• Form the final training set for Weka: run Dataset_DevType_Classification.py
• Feed the obtained CSV training set (TrainingSet_final.csv) into the Weka Explorer, then select the learning algorithm to be trained and cross-validation to evaluate its performance.