Today, our paper titled “I Choose You: Evaluating the Impact of Feature Selection on XAI Consensus for ML-NIDS” co-authored by our chair in collaboration with researchers from the University of Würzburg, Technical University of Chemnitz, and the Austrian Institute of Technology (AIT), has been presented at the 3^rd Volume of the International Workshop in Machine Learning in Networking (MaLeNe). Recent studies have shown that different explainable AI (XAI) techniques can yield inconsistent explanations, potentially caused by correlated features. In this work, we examine the hypotheses that feature selection methods can influence the level of consensus among different XAI methods. Our results show that feature selection can sometimes enhance, but also occasionally reduce, consensus, underscoring its importance for improving interpretability and consistency.

Paper: Katharina Dietz, Johannes Schleicher, Nikolas Wehner, Mehrdad Hajizadeh, Pedro Casas, Stefan Geisler, Michael Seufert, Tobias Hoßfeld. “I Choose You: Evaluating the Impact of Feature Selection on XAI Consensus for ML-NIDS”. 3^rd International Workshop in Machine Learning in Networking (MaLeNe), September 1, 2025, Illmenau, Germany.

Link to paper: Coming soon