Kubernetes application PCAPs

Overview

This dataset is a result of a version-sensitive network traffic classification framework. The framework's goal is to distinguish between different application versions based on the observed network traffic. The framework is part of the author's master thesis. The source code and thesis paper is available in the author's Github.

Dataset

The dataset consists of PCAP files from different Kubernetes application versions. The dataset also contains fingerprint comparison files in a PCAP and CSV format. Finally the dataset contains classification results between different application versions. More details in the thesis paper and source code. 

Structure

The data is stored in the following format and hierarchy:

- Root folder

  - There are multiple csv files that contain aggregated statistical information about the data, applications and results.

  - There are subfolders for each recorded application. The folders are named `-`.

    - Each subfolder contains all the recorded data for each application version deployment in PCAP format. Each recording is named `__.pcap`.

    - Each subfolder also contains a config file that can be used to recapture the recorded data.

    - Each subfolder also has a pod metadata file and an output csv file that contains a summary of the recorded PCAP files.

    - Each subfolder also contains a subfolder named `fingerprint_comparison` that contains the fingerprint comparison results and the final classification results.

      - The comparison results are stored in a csv and PCAP format. PCAP is more used for debugging and the csv file is used to generate the final `aggregated_results.csv` file which is fed to the machine learning model.

      - The final classification results are stored in the `prediction_results.csv` file.