PrivBench
收藏资源简介:
PrivBench是由深圳大学开发的一个创新的数据合成框架,专注于在保证隐私的前提下生成高质量的数据集,用于数据库性能的基准测试。该数据集利用了和积网络(SPNs)来分割和采样数据,增强了数据表示的同时确保了隐私安全。PrivBench允许用户调整SPN分割的细节和隐私设置,这对于定制隐私级别至关重要。数据集的应用领域主要是数据库性能测试,旨在解决在保护用户隐私的同时,如何生成与原始数据查询性能相近的合成数据集的问题。
PrivBench is an innovative data synthesis framework developed by Shenzhen University, dedicated to generating high-quality datasets for database performance benchmarking while guaranteeing user privacy. This framework adopts Sum-Product Networks (SPNs) to split and sample data, enhancing data representation while ensuring privacy and security. PrivBench allows users to adjust the granularity of SPN-based segmentation and privacy settings, which is critical for customizing privacy levels. Its primary application domain is database performance testing, aiming to address the challenge of generating synthetic datasets that exhibit query performance consistent with that of the original data while protecting user privacy.

- 1Privacy-Enhanced Database Synthesis for Benchmark Publishing深圳大学 · 2024年



