Differentially Private Combinatorial Cloud Auction

Cloud service providers typically provide different types of virtual machines (VMs) to cloud users with various requirements. Thanks its effectiveness and fairness, auction has been widely applied in this heterogeneous resource allocation. Recently, several strategy-proof combinatorial auction mechanisms have proposed. However, they fail protect the bid privacy from being inferred results. In article, we design a xmlns:xlink="http://www.w3.org/1999/xlink">differentially private mechanism (DPCA) address issue. Technically, employ exponential compute clearing unit price vector probability proportional corresponding revenue. We further improve reduce running time while maintaining high revenues, by computing single price, or subgroup prices at time, resulting improved DPCA-S generalized version DPCA-M, respectively. theoretically prove that our can guarantee differential privacy, approximate truthfulness Extensive experimental results demonstrate DPCA generate near-optimal revenues relatively complexity, achieve tunable trade-off between revenue time.