Several researchers have illustrated that data privacy is an important and inevitable constraint when dealing with distributed knowledge discovery. The challenge is to obtain valid results while preserving this property in each related party. In this paper, we propose a new approach based on enrichment of graphs where each party does the cluster of each entity (instance), but does nothing about the attributes (features or variables) of the other parties. Furthermore, no information is given about the clustering algorithms which provide the different partitions. Finally, experiment results are provided for validating our proposal over some known data sets.