How ivoafonsobispo/xai-cybersecurity works
Project type and purpose
• This is a research-oriented Machine Learning pipeline for an Explainable AI Intrusion Detection System (IDS).
• It is not a standard client-server application, but a data science workflow with distinct phases: datasets ingestion, pre-modeling (EDA, feature engineering), modeling (training, evaluation, saving), and explainability (feature importance, individual prediction, other XAI techniques).
File structure
• Top-level directories:
• 0_Datasets – raw input data
• 1_Pre-Modeling_Phase – data exploration, feature engineering, preprocessing
• 2_Modeling_Phase – training, evaluation, old models, saved model artifacts
• 3_Explainable_Phase – various XAI analyses (feature importance, individual predictions, other libraries)
• Many Jupyter notebooks indicate a sequential pipeline rather than microservices or MVC.
README for additional insights
• The Experimental Workflow image describes the high-level flow: ingestion → pre-modeling → modeling → explainable.
• Technologies implied: Python, Jupyter, scikit-learn, XGBoost, TensorFlow/Keras, SHAP/LIME, Fairlearn, AIF360, etc.