Privacy-Preservation for Distributed Quantized k-means Clustering

In this paper, we enhance a distributed version of the well known k-means algorithm with privacy-preservation features. While ensuring that sensitive or confidential information remains undisclosed to unauthorized entities (in our case these entities are curious nodes within the network), we maintain the desirable features of the distributed algorithm: the transmitted values are quantized, which optimizes bandwidth utilization and alleviates communication bottlenecks, and nodes possess the ability to collectively determine when to terminate the algorithm, which enables the conservation of valuable resources. We introduce a novel privacy-preserving protocol that not only preserves the privacy of a node's state, but it also ensures confidentiality about its cluster affiliation (i.e., it does not reveal to curious nodes whether a node participates in the update calculation for a specific centroid value). Moreover, we precisely characterize topological conditions that guarantee privacy preservation for individual nodes. Our distributed algorithm allows the formation of exclusive clusters within a finite time frame on any static and strongly connected directed graph.