A Task-based Personalized Privacy-Preserving Participant Selection Mechanism for Mobile Crowdsensing

Lele Zheng, Tao Zhang, Yulong Shen, Bowen Deng, Ze Tong

Abstract: Participant selection is one of the most crucial problems in mobile crowdsensing, which relies on the feature similarity between tasks and mobile users to select participants appropriately. However, since the server cannot be fully trustable, participant selection suffers some security and privacy challenges. Although there are many approaches to protect users’ privacy, the privacy of sensitive task requesters has been neglected. The untrusted server may infer private information such as the requester’s hobbies and location from the sensitive task. In addition, none of the existing privacy-preserving participant selection mechanisms can provide personalized protection considering diverse protection needs of users. In this paper, we propose a task-based personalized privacy protection participant selection mechanism that can securely select participants based on the attribute vector of the task while providing personalized privacy protection for participants. The basic idea is that the task requester and candidate users perturb their attribute vector and upload them to the server. The server performs similarity estimation on the obfuscated attribute vector to select high-quality participants. Furthermore, we propose a truth discovery algorithm for obfuscated values to improve the final result’s accuracy. We also theoretically analyze the privacy and utility of our propsoed mechanism, which outperforms the state-of-the-art solutions in both privacy protection degrees and time complexity. Experiments on synthetic datasets validate the effectiveness and efficiency of our mechanism.

External IDs:dblp:journals/monet/ZhengZSDT23