Data-Matching-Based Privacy-Preserving Statistics and Its Applications in Digital Publishing Industry

With the rapid development of digital media technology, many people prefer to read e-books over article versions. The digital publishing platform can collect and analyze massive amounts of readers’ reading information. The statistical analysis results can be regarded as the platform's digital assets based on which it provides paid services for its users. However, three privacy issues are related to readers’ reading information, users’ statistical preferences, and the platform's digital assets. This article proposes a data-matching-based privacy-preserving statistic scheme. The proposed solution combines bloom filters, secret sharing, and perturbing technologies to realize an efficient match between users’ statistical preferences and massive readers’ corresponding reading information and statistical analysis of the matching results without compromising the privacy of different parties. Besides, the proposed solution adopts an edge computing paradigm to realize the process of massive data in a divide-and-conquer parallel way. It introduces the concepts of Mirror Secret Shares and Buddy Edge Devices to virtualize the
$(m+1, m+1)$
-threshold secret sharing scheme to an
$(m+1, m+\lfloor m/2 \rfloor +2)$
-threshold secret sharing scheme for achieving good robustness without adding hardware devices. The detailed analyses show that our solution meets the defined design goals. Furthermore, the experimental results demonstrate the efficiency of the proposed work.