A Scalable Scheme for Privacy-Preserving Aggregation of Time-Series Data

摘要

Suppose that a set of multiple users uploads in every time period encrypted values of some data. The considered problem is how an untrusted data aggregator can compute the sum of all users' values but nothing more. A solution was recently given by Shi et al. (NDSS 2011). However, as advocated by the authors, the proposed encryption scheme suffers from some limitations. In particular, its usage is restricted to small plaintext spaces. This paper presents a practical scheme which, advantageously, can accommodate large plaintext spaces. Somewhat surprisingly, it comes with an efficient security reduction, regardless of the number of users. Furthermore, the proposed scheme requires a minimal number of interactions, is efficient for both encryption and decryption/aggregation and can operate in an off-line/on-line mode.