Private Stream Aggregation Revisited

Abstract: In this work, we investigate the problem of private statistical analysis in the distributed and semi-honest setting. In particular, we study properties of Private Stream Aggregation schemes, first introduced by Shi et al. \cite{2}. These are computationally secure protocols for the aggregation of data in a network and have a very small communication cost. We show that such schemes can be built upon any key-homomorphic \textit{weak} pseudo-random function. Thus, in contrast to the aforementioned work, our security definition can be achieved in the \textit{standard model}. In addition, we give a computationally efficient instantiation of this protocol based on the Decisional Diffie-Hellman problem. Moreover, we show that every mechanism which preserves $(\epsilon,\delta)$-differential privacy provides \textit{computational} $(\epsilon,\delta)$-differential privacy when it is executed through a Private Stream Aggregation scheme. Finally, we introduce a novel perturbation mechanism based on the \textit{Skellam distribution} that is suited for the distributed setting, and compare its performances with those of previous solutions.