Private and oblivious set and multiset operations

摘要

Privacy-preserving set operations are a popular research topic. Despite a large body of literature, the great majority of the available solutions are two-party protocols and expect that each participant knows her input set in the clear. In this work, we put forward a new framework for secure multi-party set and multiset operations in which the inputs can be arbitrarily partitioned among the participants, knowledge of an input (multi)set is not required for any party, and the secure set operations can be composed and can also be securely outsourced to third-party computation providers. In this framework, we construct a comprehensive suite of secure protocols for set operations and their various extensions. Our protocols are secure in the information-theoretic sense and are designed to minimize the round complexity. We then also build support for multiset operations by providing (i) a generic conversion from a multiset to a set, which makes the protocols for set operations applicable to multisets and (ii) direct instantiations of multiset operations of improved performance. All of our protocols have communication and computation complexity of and logarithmic round complexity for sets or multisets of size m, which compares favorably with prior work. Practicality of our solutions is shown through experimental results, and novel optimizations based on set compaction allow us to improve performance of our protocols in practice. Our protocols are secure in both semi-honest and malicious security models.