Explainable AI (XAI) for Cybersecurity: Intrusion Detection System (IDS)

Introduction

The increasing use of Machine Learning models in Cybersecurity applications has raised concerns about their interpretability and transparency. This research project aims to develop explainable Machine Learning-Based Intrusion Detection System (IDS) models and improve their performance.

Project Summary

This research project, funded by FCT - Foundation for Science and Technology UIDB/04524/2020, is developed by CIIC - Computer Science and Communication Research Centre - Polytechnic of Leiria.

Repository Structure

• Datasets:
  • Contains the datasets used for creating the Machine Learning IDS models.
• Pre-Modeling Phase:
  • Includes notebooks and scripts for the pre-modeling phase.
  • Exploratory Data Analysis (EDA): Notebooks for initial data exploration and visualization.
  • Feature Understanding: Scripts for analyzing and understanding the features in the datasets.
  • Feature Engineering: Notebooks for creating new features or modifying existing ones to improve model performance.
  • Pre-Processing: Scripts for data cleaning, normalization, and preparation before modeling.
• Modeling Phase:
  • Contains scripts and notebooks for the modeling phase.
  • Model Creation: Scripts for training various Machine Learning models.
  • Model Evaluation: Notebooks for assessing the performance of different models.
  • Model Saving: Scripts for saving the trained models for future use.
• Explainable Phase:
  • Includes notebooks with explainability libraries and tools.
  • Feature Importance: Notebooks for determining the importance of each feature in the models.
  • Individual Predictions: Tools for explaining specific predictions made by the models.
  • Other Libraries: Additional libraries and methods for enhancing the interpretability of the models.

Experimental Workflow

Getting Involved

• Contributing: Contact the authors with questions or contributions: Ivo Bispo ([email protected]).