Verifying the Correctness of Analytic Query Results (Extended Abstract)

Abstract

This research studies the problem of enabling users to verify that the results of analytical queries such as top k they receive from a potentially untrustworthy cloud are indeed correct. Existing work shows that it is possible for a data owner to create an authentication data structure (ADS) by which a cloud can build a verification object (VO) to prove the correctness of a query result. The current technique, however, has largely ignored the computation cost in VO construction and query result verification. In this paper, we extend and integrate Intersection tree (I-tree) and Merkle hash-tree (MH-tree) to develop a new ADS called Intersection Function Merkle Hash-tree (IFMH-tree). We propose two versions of the IFMH-tree, one-signature and multi-signature, and study their performance in supporting three representative types of analytic queries, including top-k, range, and KNN queries. Our results show that the new technique outperforms the existing solution to a large extent.