Owner-centric control is a widely adopted method for easing owners' concerns over data abuses and motivating them to share their data out to gain collective knowledge. However, while many control enforcement techniques have been proposed, privacy threats due to the metadata leakage therein are largely neglected in existing works. Unfortunately, a sophisticated attacker can infer very sensitive information based on either owners' data control policies or their analytic task participation histories (e.g., participating in a mental illness or cancer study can reveal their health conditions). To address this problem, we introduce $\mathsf{\text{Vizard}}$ , a metadata-hiding analytic system that enables privacy-hardened and enforceable control for owners. $\mathsf{\text{Vizard}}$ is built with a tailored suite of lightweight cryptographic tools and designs that help us efficiently handle analytic queries over encrypted data streams coming in real-time (like heart rates). We propose extension designs to further enable advanced owner-centric controls (with AND, OR, NOT operators) and provide owners with release control to additionally  regulate how the result should be protected before deliveries. We develop a prototype of $\mathsf{\text{Vizard}}$ that is interfaced with Apache Kafka, and the evaluation results demonstrate the practicality of $\mathsf{\text{Vizard}}$ for large-scale and metadata-hiding analytics over data streams.

Based on the following work, Chengjun Cai, Yichen Zang, Cong Wang, Xiaohua Jia, Qian Wang, 'Vizard: A Metadata-hiding Data Analytic System with End-to-End Policy Controls', ACM CCS 2022.

About the speaker