Privacy-Preserving Multimedia Big Data Aggregation in Large-Scale Wireless Sensor Networks

Abstract

To preserve the privacy of multimedia big data and achieve the efficient data aggregation in wireless multimedia sensor networks (WMSNs), a distributed compressed sensing--based privacy-preserving data aggregation (DCSPDA) approach is proposed in this article. First, in this approach, the original multimedia sensor data are compressed and measured by distributed compressed sensing (DCS) and the compressed data measurements are uploaded to the sink, by which the inherent characteristics between sensor data can be obtained. Second, the original multimedia data are jointly recovered and the common and innovation sparse components are obtained through solving the optimization problem and linear equations at the sink. Third, through least squares support vector machine (LSSVM) learning of the sparse components, the sparse position configuration can be determined and disseminated for each node to conduct the privacy-preserving data configuration. After receiving the configuration message, original multimedia sensor data are accordingly customized, compressed, and measured by the common measurement matrix, aggregated at the cluster heads, and transmitted to the sink. Finally, the aggregated multimedia sensor data are recovered by the sink according to the data configuration to achieve the privacy-preserving data aggregation and transmission. Our comparative simulation results validate the efficiency and scalability of DCSPDA and demonstrate that the proposed approach can effectively reduce the communication overheads and provide reliable privacy-preserving with low computational complexity for WMSNs.