Secure distributed framework for achieving ε-differential privacy

摘要: Privacy-preserving data publishing addresses the problem of disclosing sensitive when mining for useful information. Among existing privacy models, e-differential provides one strongest guarantees. In this paper, we address private where is horizontally divided among two parties over same set attributes. particular, present first generalization-based algorithm differentially release horizontally-partitioned between in semi-honest adversary model. The generalization correctly releases differentially-private and protects each party according to definition secure multi-party computation. To achieve this, a two-party protocol exponential mechanism. This can be used as subprotocol by any other that requires mechanism distributed setting. Experimental results on real-life suggest proposed effectively preserve information task.