Demonstration of DPClustX: Differentially Private Explanations for Clusters

We present DPClustX, a framework designed to generate Differential Privacy (DP)-compliant histogram-based explanations for black-box clustering results. Interpreting clustering results is challenging even without privacy constraints, and the challenge is amplified under DP, as analysts receive noisy query responses. By addressing these challenges, DPClustX enables the selection of high-quality explaining attributes and the generation of informative histograms that balance privacy and utility. Compatible with any DP-clustering algorithm, our framework provides explanations that uncover significant patterns even under strict privacy constraints. This demonstration highlights DPClustX's capabilities using real datasets, illustrating its practical utility in sensitive data analysis and its potential for enhancing the interpretability of clustering methods.