Balancing GAN Algorithm for Cyberattack Datasets

Faculty Advisors:

• Dr. Qu, Chenqi

• Dr. Prasad, Calyam

• Dr. Mazzola Luca

Mentor:

• Kevin Kostage: Deep Learning Engineer, End-to-End Engineer, Cloud-Network Specialist

Undergraduate Researchers:

• Please Refer to contributions and contributors.md

• Most contributions were made in Researching and Exploring Articles such as discovering the CICIOT2023 dataset. Weekly Homework assignments were optimizing or developing their version of functions and processes such as loading and processing tabular datasets for models, model structures, compiling models, and training. We also had workshops configuring virtual environments with Anaconda and CUDA.

• Students contributed 10 hours a week between March 1st - April 24th. Awarded Best Class Presentation.

• Contributions from the team's HW assignments were submitted from Keko787.

Publication

Enhancing Autonomous Intrusion Detection System with Generative Adversarial Networks:

• Paper: https://www.researchgate.net/publication/384221099_Enhancing_Autonomous_Intrusion_Detection_System_with_Generative_Adversarial_Networks
• Github Repo: https://github.com/Keko787/Generating-a-Balanced-IoT-Cyber-Attack-Dataset-with-GAN-COIL-Collaboration-

Previous Works

Enhancing Drone Video Analytics Security Management using an AERPAW Testbed:

• Paper: https://ieeexplore.ieee.org/document/10620812

Posters

Purpose

• To develop a Generative Adversarial Network to classify fake traffic in network traffic feeds and augment training data with realist synthetic data to balance the dataset and provide robust data.
• To develop and discover the most effective and efficient GAN variant
• Understand how to process cybersecurity data to train a DNN model

Models

• CGAN
• CWGANGP
• CTGAN
• WGAN

Steps to run

• Clone Repo

• Download the CICIOT dataset extract and put it into the repo directory under a folder as archive or parent_dir/archive

• run the .py files in parent dir