Differentially Private Data Aggregation with Optimal Utility

We present PrivaDA, a tool for computing differentially private statistics about user data, which builds on advanced SPMC schemes. PrivaDA supports a variety of queries and perturbation mechanisms. At the core of this approach lies the first generic technique to generate noise in a distributed manner, while maintaining optimal utility and without suffering from the answer pollution problem.

Author(s):

Fabienne Eigner    
Saarland University, CISPA
Germany

Aniket Kate    
MMCI, Saarland University
Germany

Matteo Maffei    
Saarland University, CISPA
Germany

Francesca Pampaloni    
IMT Lucca
Italy

Ivan Pryvalov    
MMCI, Saarland University
Germany