Datasets used for the experiment

Training and testing data:

We have taken Aposemat IoT-23 and Edge-IIoTset datasets for training and testing the machine learning models. The IoT-23 dataset is semi-structured logging information of the packets labeled with malicious and benign IoT network traffic. The dataset was originally created by Avast AIC laboratory collecting from different IoT devices. Another dataset Edge-IIoTset is already pre-processed. However, we have carried out further pre-processing and cleaning steps.

The original dataset path for the project:

DATA_DIR = 'data/iot_23_datasets_small/'

The generated csv files (raw without filtering) generated from original textual files:

DATA_DIR = 'data/processed/<processed_data_file>.csv'

The processed data path after preprocessing/filtering and feature selection:

DATA_DIR = 'data/processed/<processed_data_file>.csv'

Prediction data:

The predicting data is the locally running network traffic data from the IoT test cases of nodes in local smart environments. Configure predict.yaml file to retrieve the prediction data and apply the trained model. The network traffic data using zeek can be given as providing log_file parameter value:

log_file: data/predict/conn.log