DNF-BLPP: An Effective Deep Neuro-Fuzzy Based Bilateral Location Privacy-Preserving Scheme for Service in Spatiotemporal Crowdsourcing

Abstract

Mobile Crowd Sensing (MCS) is an emerging paradigm that constructs various services by recruiting massive workers in edge networks to sense data. However, ensuring Quality-of-Service (QoS) while preserving bilateral location privacy remains a critical challenge for effective service provisioning in the context of spatiotemporal crowdsourcing. Previous studies have achieved privacy preservation through location obfuscation, which has the problem of low task-worker matching rate, and data loss adversely affected QoS. To tackle this issue, we propose a Deep Neuro-Fuzzy based Bilateral Location Privacy-Preserving (DNF-BLPP) scheme to construct services in spatiotemporal crowdsourcing. In this article, we first present a novel obfuscation strategy that obfuscates the location of each task and worker respectively to ${\rm{\lambda }}$ locations with the highest correlation, ensuring privacy preservation while providing assurance for accurate data recovery. Then, we further introduce a deep neural-fuzzy approach to solve the worker selection problem under obfuscated locations. Based on that, a Non-negative Constraint Matrix Factorization algorithm is employed to accurately impute missing data based on time-space correlation. Theoretical analysis and extensive simulations show that the proposed scheme has a strong ability to protect location privacy, and is better than the state-of-the-art schemes in performance indicators such as task-worker matching rate, QoS and computational efficiency.