多源数据融合场景下的对抗式隐私洞察靶向保护技术研究

多源数据安全融合技术是企业间合作共赢、我国共享经济发展的核心助力。如何在数据融合的过程中感知隐私演化的态势、将隐私保护效能与数据可用性等多维需求纳入一致性考量，在信息安全可控下实现信息的个性化共享，已成为数据多方共享实践不可回避的重要课题。本项目指出已有隐私保护技术在隐私演化洞察、隐私靶向保护、隐私对抗保护三方面的不足，并以此为突破点，拟开展多源数据融合中隐私洞察计算基础理论与量化评估、数据隐私性与可用性的内在联系及隐私靶向技术对其的优化调节、面向多源数据增量动态融合的对抗式隐私保护攻防机制与算法三方面关键技术的研究。随着上述关键技术的突破，将构建以医疗数据安全融合为示范的验证系统，并基于真实的医疗大数据对关键技术进行验证。本项目旨在为企业间数据安全共享提供必要的基本理论和技术支撑。

The privacy-preserving multiparty data fusion technology is the core driving force for win-win cooperation among enterprises and the development of China's sharing economy. How to perceive the situation of privacy evolution in the process of data fusion, integrate the multidimensional requirements of privacy protection efficiency and data availability into the consistency consideration, and realize the personalized sharing of information under the control of information security has become an important issue that can not be avoided in the practice of data multiparty sharing..This project points out the shortcomings of the existing privacy protection technologies in privacy insight computing, privacy-targeted protection, and adversarial privacy protection. To address these problems, we focus of three important issues including the basic theory and quantitative evaluation of privacy insight computing, the optimization and adjustment of data privacy and usability by the privacy-targeted protection technology, and the adversarial privacy protection mechanism and algorithm for incremental dynamic fusion of multi-source data. We intent to make several breakthroughs in theoretical models and key algorithms. Further, we plan to build a verification system with medical data security fusion as a demonstration, and verify the effectiveness of our technologies based on real medical big data..The project aims to provide fundamental theoretical support as well as key technology innovations to the development of secure and effective data-sharing techniques among enterprises.