UC5 : Machine Learning vs DDoS/en

Développement du Use Case pédagogique "Machine Learning vs Attaque DDoS" dans le cadre du GT IA et Cyber

Catégorie : Commun Statut : Production 1 : Idée - 2 : Prototype - 3 : Validation - 4 : Production

Overview

The use case focuses on vast varieties of intrusions and attack activities of the network traffic. We propose three Dataset ready for exploration and modeling.

• [CSE-CIC-IDS2018Datasets](https://www.unb.ca/cic/datasets/ids-2018.html#:~:text=In%20CSE-CIC-IDS2018%20dataset%2C%20we%20use%20the%20notion%20of,human%20operators%20to%20generate%20events%20on%20the%20network)

• [UNSW-NB15Datasets](https://research.unsw.edu.au/projects/unsw-nb15-dataset)

• [USB-IDS Datasets](http://idsdata.ding.unisannio.it/index.html)

For each cyber security dataset we propose data analysis, data standardization an modeling notebooks. In some notebooks standardization step is include in the modeling notebook

• Authors: Christian Maréchal
• Keywords: Supervised Machine Learning, Half-Supervised Machine Learning, Clustering, Data Standardization

Plan of Study

• Data Analysis Notebook --\> Standardization Notebook --\> Modeling Notebook

To make things easier each notebook can be run independently.

Dataset Cse cic ids

Data

cleaned_ids2018S_test.cs
cleaned_ids2018S_train.csv

Notebooks

Notebook	Data Science step
Cyber_cse-cic-ids_analysis.ipynb	Data exploration
Cyber_cse-cic-ids_model.ipynb	Standardization and Half-Supervised Autoencoder model

Dataset Unsw

Data

NUSW-NB15_features.utf8.csv
UNSW_NB15S_test.csv
UNSW_NB15S_train.csv
UNSW-NB15_1.csv
UNSW-NB15_2.csv
UNSW-NB15_3.csv
UNSW-NB15_4.csv

Notebooks

Notebook	Data Science step
Cyber_unsw_analysis.ipynb	Data exploration
Cyber_unsw_analysisGmm.ipynb	data exploration for GMM clustering
Cyber_unsw_standardization.ipynb	data standardization
Cyber_unsw_autoencoder.ipynb	Binary classifier study. Half-Supervised Autoencoder modeling, we tested:
	-logistic regression
	-Autoencoder Inria like
	-Autoencoder single layer
	-Autoencoder multi layers
Cyber_unsw_complete_analysis.ipynb	data exploration
Cyber_unsw_model.ipynb	Data Supervised model, to classify attacks of different kinds, we tested:
	-Random Forest Classifier (rfc)
	-Support Vector Classification (svm)
	-Multi-Layer Perceptron (mlp)
	-Artificial Neural Network (ann)
	-eXtreme Gradient Boosting (xgb)
	-Convolutional Neural Network (cnn)

Dataset USB IDS

Data

USB-IDS-1S-TEST.csv
USB-IDS-1S-TRAIN.csv
USB-IDS-1S-VALIDATION.csv

Notebooks

Notebook	Data Science step
Cyber_USB-IDS_analysis.ipynb	Data exploration

Notebooks du use case

Retrouvez tous les éléments du Use Case sur le GitLab du Campus Cyber :https://gitlab.com/campuscyber/gt-ia-et-cyber/-/tree/main/UC5%20Cyber%20Attack%20Use%20Case%20-%20Machine%20Learning%20Vs%20DDoS?ref_type=heads

Groupe de travail

IA et cybersécurité