DC604 Monthly Meetup

​Voting for Transparency: Ensemble Models and Explainable AI in Intrusion Detection with Humberto Goncalves

​This talk is based on Humberto's research paper, accepted at the IEEE SOLI 2025 Conference, which addresses how intrusion detection can be made both more effective and more transparent for security professionals. Ensemble Machine Learning methods and Explainable AI (XAI) are explored as ways to enhance detection capabilities without sacrificing interpretability. The session will introduce intrusion detection systems and their challenges, then show how ML models (XGBoost and CatBoost) were combined through Voting Strategies to boost performance, and how SHAP, an XAI technique, was applied to clarify model decisions.

Humberto will cover key concepts such as hard vs. soft voting, performance metrics, confusion matrices, and model interpretability, with these ideas presented in terms accessible to newcomers and experienced practitioners. The talk will conclude with a discussion of performance trade-offs, research limitations, and the implications of combining ML models, Voting Strategies, and XAI in real-world SOC environments.

The goal is to equip the community with a clear understanding and actionable strategies for strengthening modern intrusion detection systems through ensemble learning and explainability.