Efficient Data Trading for Stable and Privacy Preserving Histograms in Internet of Things

Abstract

Internet of Thing (IoT) systems provide novel opportunities for data acquisition, where sensing devices can flexibly collect and trade data with data brokers. A data broker may conduct sophisticated analysis on the collected data and further exchange statistics like histograms with data requestors. Considering the supply-demand correlations, several pivotal factors must be jointly treated including communication bandwidths, data utilities, privacy issues, total budget, etc. Unfortunately, the current efforts mainly apply the crowdsensing strategies during data trading and overlook the subsequent processing and analysis of the collected data. Therefore, this paper proposes a novel framework for efficient data trading in IoT systems throughout the data collection and data processing phases. In the framework, data contributors can flexibly arrive and departure from the monitored area in heterogeneous time slots. The incentives for data trading are correlated with data volume, channel condition, and privacy issues of each contributor. Meanwhile, a data broker samples partial sensing data and aggregates approximate histograms for data requestors. The objective is to minimize the total budget for data trading. First, the theoretical bound on the necessary budget for histograms with a given accuracy is proved. Then two algorithms are proposed for efficient data trading among data contributors, based on whether the behaviors of data contributors are known in advance. Both algorithms are analyzed and the corresponding guarantees on performance are discussed. Finally, the extensive evaluation results validate the advancement of the proposed algorithms.