Privacy-Preserving Task Allocation via Local Differential Privacy in Vehicular Crowdsensing

In vehicular crowdsensing (VCS), task allocation necessitates a fundamental trade-off between service quality and user privacy. A central challenge in VCS is that, in order to guarantee spatio-temporal coverage by assigning tasks to suitable workers, the service provider inevitably requires their trajectories. Since the service provider is not fully trusted, this requirement raises significant privacy concerns. To address this problem, we propose an enhanced Optimized Unary Encoding (OUE)-based differential private task allocation (ODTA) scheme in VCS. Unlike traditional task allocation schemes, we employ local differential privacy to achieve privacy-preserving task allocation, which can enable both requesters and workers to locally encode and perturb task regions and predicted trajectories, respectively, without any trusted third party. The service provider first collects the perturbed region and trajectory encodings from both requesters and workers. It then computes disparities between the task region and each worker trajectory using a model based on Hamming distance to produce an approximately optimal matching between task and workers. Security analysis shows that ODTA satisfies $\varepsilon$-differential privacy. Extensive experiments on a simulation dataset based on real-world road networks demonstrate that ODTA maintains high communication and computational efficiency while achieving substantial spatiotemporal coverage.