Characterizing and Optimizing Differentially-Private Techniques for High-Utility, Privacy-Preserving Internet-of-Vehicles

Yicun Duan, Junyu Liu, Xiaoxing Ming, Wangkai Jin, Zilin Song, Xiangjun Peng

Abstract: Recent developments of advanced Human-Vehicle Interactions rely on Internet-of-Vehicles (IoV), to achieve large-scale communications and synchronizations of data in practice. IoV is highly similar to a distributed system, where each vehicle is considered as a node and all nodes are grouped with a centralized server. In this manner, concerns of data privacy are rising, since privacy leak possibly occurs when all vehicles collect, process and share personal statistics (e.g. driver’s heart rate, skin conductance and etc.). Therefore, it’s important to understand how to efficiently apply modern privacy-preserving techniques on IoV. In this work, we first present a comprehensive study to characterize modern privacy-preserving techniques for IoV, and then propose a Differential Privacy(DP) privacy-protection framework specialized for unique characteristics of IoV. Our characterization focuses on DP, a representative set of mathematically-guaranteed mechanisms for both privacy-preserving processing and sharing of sensitive data. It demystifies the tradeoffs of deploying DP techniques, in terms of service quality and privacy-preserving effects. The lessons learned from characterization reveal the importance of data utility in DP-protected IoV and motivate us to examine new opportunities. To better balance tradeoffs and improve service quality, we introduce HUT, for high-utility, batched queries under DP-protection on IoV. We quantitatively examine the benefits of HUT, and experimentally show that, in an IoV context, HUT can reduce information loss by 95.69% while enabling strong mathematically-guaranteed protection over sensitive data. Based on our characterization and optimizations, we identify key challenges and opportunities for future studies, to enable privacy-preserving IoV with low service quality degradation.