P4P: Practical Large-Scale Privacy-Preserving Distributed Computation Robust against Malicious Users

In this paper we introduce a framework for privacypreserving distributed computation that is practical for many real-world applications. The framework is called Peers for Privacy (P4P) and features a novel heterogeneous architecture and a number of efficient tools for performing private computation and ensuring security at large scale. It maintains the following properties: (1) Provably strong privacy; (2) Adequate efficiency at reasonably large scale; and (3) Robustness against realistic adversaries. The framework gains its practicality by decomposing data mining algorithms into a sequence of vector addition steps that can be privately evaluated using a new verifiable secret sharing (VSS) scheme over small field (e.g., 32 or 64 bits), which has the same cost as regular, non-private arithmetic. This paradigm supports a large number of statistical learning algorithms including SVD, PCA, k-means, ID3, EM-based machine learning algorithms, etc., and all algorithms in the statistical query model [36]. As a concrete example, we show how singular value decomposition (SVD), which is an extremely useful algorithm and the core of many data mining tasks, can be done efficiently with privacy in P4P. Using real-world data and actual implementation we demonstrate that P4P is orders of magnitude faster than existing solutions.