Local Privacy-Preserving Dynamic Worker Locations in Spatial Crowdsourcing

An outsourcing service named spatial crowdsourcing (SC) becomes popular, whereby the SC-server allocates nearby tasks to the workers based on the outsourced task and worker locations. Exposing real locations can cause serious privacy leakage. However, traditional differential privacy (DP) and encryption methods do not consider the dynamic worker location and correlation privacy. Here, a Local DP-based dynamic worker location protection (LDPDW) scheme is proposed to achieve high-quality task allocation and locally protect the correlation and location privacy of dynamic workers. Specifically, LDPDW generates noisy high correlated graph classes and obfuscates the worker locations in a static case by adopting a LDP-based correlation graph (LDPCG) algorithm and distance score-based LDP (DSLDP) algorithm, thereby achieving controlled noise addition and ensuring the correlation and location privacy. To support the privacy-preserving dynamic locations, a dynamic correlation graph-based location obfuscation (DCGLO) algorithm is proposed to allocate reasonable privacy budget
$\epsilon $
, which ensures the data utility. Finally, a linear acceptance model-based task allocation (LAMTA) algorithm is used to allocate tasks to the workers with high acceptance rates. Privacy analysis and the extensive experimental results show that our LDPDW scheme follows
$\epsilon $
-LDP while allocating tasks with high data utility.