Achieving Efficient and Privacy-Preserving k-NN Query for Outsourced eHealthcare Data.

The boom of Internet of Things devices promotes huge volumes of eHealthcare data will be collected and aggregated at eHealthcare provider. With the help of these health data, eHealthcare provider can offer reliable data service (e.g., k-NN query) to doctors for better diagnosis. However, the IT facility in the eHealthcare provider is incompetent with the huge volumes of eHealthcare data, so one popular solution is to deploy a powerful cloud and appoint the cloud to execute the k-NN query service. In this case, since the eHealthcare data are very sensitive yet cloud servers are not fully trusted, directly executing the k-NN query service in the cloud inevitably incurs privacy challenges. Apart from the privacy issues, efficiency issues also need to be taken into consideration because achieving privacy requirement will incur additional computational cost. However, existing focuses on k-NN query do not (fully) consider the data privacy or are inefficient. For instance, the best computational complexity of k-NN query over encrypted eHealthcare data in the cloud is as large as [Formula: see text], where N is the total number of data. In this paper, aiming at addressing the privacy and efficiency challenges, we design an efficient and privacy-preserving k-NN query scheme for encrypted outsourced eHealthcare data. Our proposed scheme is characterized by integrating the k d-tree with the homomorphic encryption technique for efficient storing encrypted data in the cloud and processing privacy-preserving k-NN query over encrypted data. Compared with existing works, our proposed scheme is more efficient in terms of privacy-preserving k-NN query. Specifically, our proposed scheme can achieve k-NN computation over encrypted data with [Formula: see text] computational complexity, where l and N respectively denote the data dimension and the total number of data. In addition, detailed security analysis shows that our proposed scheme is really privacy-preserving under our security model and performance evaluation also indicates that our proposed scheme is indeed efficient in terms of computational cost.