Graph-based Clustering under Differential Privacy

Rafael Pinot (1, 2) , Anne Morvan (1, 2) , Florian Yger (1) , Cedric Gouy-Pailler (2) , Jamal Atif (1)

1 LAMSADE - Laboratoire d'analyse et modélisation de systèmes pour l'aide à la décision
2 LI3A (CEA, LIST) - Intelligence Artificielle et Apprentissage Automatique (CEA, LIST)

In this paper, we present the first differentially private clustering method for arbitrary-shaped node clusters in a graph. This algorithm takes as input only an approximate Minimum Spanning Tree (MST) T released under weight differential privacy constraints from the graph. Then, the underlying nonconvex clustering partition is successfully recovered from cutting optimal cuts on T. As opposed to existing methods, our algorithm is theoretically well-motivated. Experiments support our theoretical findings.

Mots clés: graph; statistical analysis; Minimum Spanning Tree; weight differential privacy constraint; nonconvex clustering partition; clustering; arbitrary-shaped node cluster; Artificial inteligence

Domaines: Analyse de données, Statistiques et Probabilités [physics.data-an]; Intelligence artificielle [cs.AI]

Rafael Pinot, Anne Morvan, Florian Yger, Cedric Gouy-Pailler, Jamal Atif. Graph-based Clustering under Differential Privacy. Conference on Uncertainty in Artificial Intelligence (UAI 2018), Aug 2018, Monterey, California, United States. pp.329-338. ⟨hal-02170699⟩