关键词: 隐私计算, 多方安全计算, 联邦学习, 数据生产要素

Abstract: With the rapid development of digital economy, data security has become an important challenge. How to release data value while ensuring the data privacy is a key issue in the this digital age. As a logical combination of secure multi-party computation and federated learning, privacy preserving computing has widely attracted the attentions from many tech providers and data holders. Basing on the modern cryptography techniques such as secret sharing, oblivious transfer, and garbled circuit, privacy preserving computing can enable efficient cross-domain data sharing and data fusion in a secure manner. To date, the adoption of privacy-preserving computing has taken place in many realworld scenarios, for example, finance, telecommunication, health care, and government affairs. In this paper, basing on the basic conceptions of privacy preserving computing, a brief analysis of key technologies of secure multi-parity computation and federated learning will be presented. Several typical applications in the data security scenarios will be provided. Furthermore, for addressing the interconnection problem between heterogeneous platforms, two practical approaches relying on middleware and blockchain will be discussed in the latter part of this paper.

Key words: privacy preserving computing, secure multi-party computation, federated learning, data production factor