Providing appropriate monetary incentives for participants is vital for crowdsensing to encourage their participation. Among all outstanding incentive mechanisms, posted pricing has been widely adopted because it is easy to implement and naturally achieves truthfulness and fairness. However, existing schemes either lack of privacy protection for the sensing data of participants, or fail to consider the diversity of sensing quality in crowdsensing systems. To address these critical problems, we propose a privacy-preserving dynamic pricing mechanism for robust crowdsensing, which only needs to spend a small amount of total payments to recruit a group of mobile users with reasonable sensing quality while protecting the sensing data privacy of each participant. Specifically, we first design an efficient secure aggregation algorithm through which the platform can compute the sum of sensing data from participants without learning each participant’s individual data. Then, we employ the aggregation algorithm to design a secure quality assessment algorithm to obtain the sensing quality levels of participants. Finally, according to the varying quality levels, we develop a model-free reinforcement learning based approach to optimize pricing policy to achieve lower total payments and robustness requirement. Through privacy analysis and extensive experiments, we demonstrate the effectiveness and efficiency of the proposed mechanism.

为参与者提供适当的金钱激励措施对于鼓励他们参与的群众感觉至关重要。在所有出色的激励机制中，公开定价已被广泛采用，因为它易于实施且自然实现了真实性和公平性。然而，现有的方案要么缺乏针对参与者的感测数据的隐私保护，要么没有考虑人群感测系统中感测质量的多样性。为了解决这些关键问题，我们提出了一种保护隐私的动态定价机制，以实现稳健的人群感知，该机制只需要花费少量的总费用即可招募一组具有合理感知质量的移动用户，同时保护每个参与者的感知数据隐私。特别，我们首先设计一种有效的安全聚合算法，该平台可通过该算法计算参与者的感测数据之和，而无需学习每个参与者的个人数据。然后，我们采用聚合算法设计一种安全的质量评估算法，以获得参与者的感知质量水平。最后，根据质量水平的变化，我们开发了一种基于无模型强化学习的方法来优化定价策略，以实现较低的总付款额和稳健性要求。通过隐私分析和广泛的实验，我们证明了该机制的有效性和效率。我们采用聚合算法设计安全质量评估算法，以获得参与者的感知质量水平。最后，根据质量水平的变化，我们开发了一种基于无模型强化学习的方法来优化定价策略，以实现较低的总付款额和稳健性要求。通过隐私分析和广泛的实验，我们证明了该机制的有效性和效率。我们采用聚合算法设计安全质量评估算法，以获得参与者的感知质量水平。最后，根据质量水平的变化，我们开发了一种基于无模型强化学习的方法来优化定价策略，以实现较低的总付款额和稳健性要求。通过隐私分析和广泛的实验，我们证明了该机制的有效性和效率。