Privacy-Preserving Clustering: A New Approach Based on Invariant Order Encryption

Digital Object Identifier 10.32754/JMT.2020.2.10 65 1Abstract—Cloud computing is increasingly used. One main use of cloud computing is the running of a machine learning algorithm. Due to the large amount of data required for these algorithms, they can no longer be run on personal computers. Uploading personal data to the cloud automatically raises the issues of confidentiality of this data. In this paper, we show through a series of experiments that an order-preserving encryption algorithm can be applied to guarantee the confidentiality of the input of two well-known clustering algorithms: K-Means and DBSCAN. We show that K-Means can be modified to be applied over the encrypted data. We also proposed a slight improvement to an order-preserving encryption scheme to ensure that it is randomized, therefore increasing its security level. Finally, after studying the performance of clustering algorithms over encrypted data we show a practical application of this idea, namely the color reduction over an encrypted image.